No-Code AI Agent Tutorial | Step-by-Step Guide & Transcript

[00:02] read as read as much of the introductions as possible. Okay, we have someone Chris from UAE, Abinina from Kolkata, uh Ankit who not mentioned where you are from but that's okay. Amin Kumar, uh

[00:19] from but that's okay. Amin Kumar, uh Clinton from South Africa, Javi Trapati. Okay, that's Nalini from Bangalore. Okay, an Amika from Pune. Rakkesh from Chennai, Sidra from Pakistan, Satya from Chennai

[00:34] Sidra from Pakistan, Satya from Chennai again. Okay. Hey uh someone saying that my um voice is echoing. I will try to fix that as soon as possible. And we are also live on LinkedIn and YouTube and I'm sure that there will be more people

[00:50] from all over the world joining us. Uh once again a very very warm welcome to once again a very very warm welcome to everyone. Um oh my god. In just a few seconds we have more than I think 200 messages that have come which is which

[01:04] is just amazing. I hope um that you know the same excitement and the energy stays uh throughout the session um and we are very um you know we're very eager to get started and looking forward to take this um ahead. So before we begin I just want

[01:21] to say one thing. whatever made you um you know click on register for this workshop um register for this webinar um I think it was a very very good instinct this isn't a webinar where you just watch the slides and take notes um there

[01:36] is no script running in the background um there's nothing pre-recorded nothing um there's nothing pre-recorded nothing staged um so over the next few hours you're going to watch a real AI agent get built and deployed live using

[01:48] nothing but plain English prompts um there's no coding as we have assured you in the workshop title itself that we are going to build a no code AI agent and if amazing experts on your laptop or desktop then you're walking out with an

[02:05] AI agent you've built yourself and I'm sure that's very exciting so we will um get started um okay there's someone saying that um you didn't receive a recording um from the previous session we're very sorry that that got missed

[02:20] out we usually take a poll at the end of the webinar to um you know understand who um who all required uh the recording. Maybe something got missed in the poll. I have your email ID um and I will ensure that you receive the

[02:36] recording of uh the webinar. Okay. All right. Um a few ground rules before we begin. Uh if you have any questions related to you know um your career in AI or any AI tool or concept, please do put them in the Q&A box and not in the chat

[02:49] box. The chat is for engagement and quick responses and the activities that we will run tonight. Q&A is where we will pull questions from during the dedicated Q&A segment in the end. And please keep um the chat relevant to

[03:03] what's happening on the screen. We have a huge room tonight um and it helps everyone when the conversation stays very focused. And please note that if there are a lot of pings coming in, we will keep the chat only visible to the

[03:16] host and the expert. so it does not keep popping up and affect everyone's workshop experience. So um in a few seconds um I will be enabling this

[03:28] option where only um the expert and I can see your chat. Okay. All right. So regarding audio and video I think um you know I am audible to everyone here.

[03:41] Please do drop a yes in the chat please. Um we tested everything thoroughly before going live. So if something looks or sounds off on your um you know on with your system. So please do check your connection, please check your

[03:54] your connection, please check your device or um just try rejoining. And finally, we know that a lot of you would be interested to receive a workshop um uh certificate um at the end of uh the session. So please do note that we will

[04:08] be launching a poll um specifically to take um you know your entries for the certificate. Uh so please ensure to participate in that poll and if you miss that um then it will be difficult for us to take any manual submissions or

[04:23] do ensure to enter your full name in the poll and you will be receiving the recording of the session as well as the um you know uh workshop certificate um by tomorrow and if you don't participate in the poll then there are chances that

[04:39] it might get missed out. So please do uh stay tuned and um yes participate in the stay tuned and um yes participate in the poll here. All right. So quick show of hands who's been to a simply learn webinar or workshop before. Uh we heard

[04:52] from someone already um that they were part of the last session. Um so you know drop first time in the chat box if you have not attended any of our sessions before. Let me repeat you can drop um you can type in send first time in the

[05:05] chat box if you have not attended any session before or you can drop the number of webinars you've attended before like two three four five um as well if there are any repeat

[05:18] attendees here you can type okay there's someone saying three wow okay someone with two someone again with three okay great okay all All right. All right. A lot of you are here for the first time

[05:34] lot of you are here for the first time and um so um you know uh we're very glad to have you here either way whether you're here for the first time or for multiple times. It's just amazing to have such a lively bunch of audience. So

[05:46] buildathon as we call it is brought to you by Simply Learn, let me just take a minute to tell you um you know about us and um what we do. We have been uh doing this for over a decade now. Training more than 15 million uh professionals

[06:02] across 150 plus countries all built around learning that's meant to change what you can actually do at work and not just what you know. Lers uh learners just what you know. Lers uh learners consistently rate us 4.8 out of five.

[06:15] And here is a very impressive fact. The average reported salary hike after average reported salary hike after completing a program sits above 50%. And um while today's workshop is specific to agentic agentic AI, we also run learning

[06:29] pathways and certificate programs across generative AI, data science, project management, cyber security, business and leadership and quite a lot more. And um yes, one more thing before we get into the workshop content um which is our

[06:43] partnerships. We work with some very big names in the industry as you can see on this slide. We have IITM institutions, globally ranked universities and on the corporate side companies like Microsoft, Google, IBM, AWS and a lot more. So

[06:59] these partnerships shape what actually gets taught to you in every single program and in every single course and you will see that the credibility flows you will see that the credibility flows um into today's workshop as well. Okay.

[07:11] All right. So with that we will uh get started with the workshop content and I can see let me just address a couple of questions that have come our way. Okay a few of you are confused about the poll. I guess I did mention the poll but

[07:26] please note that it will be launched at the end of the webinar. There is no poll when the poll is live. You would actually be able to see it on the screen yourself. um whether you've joined on your phone or your laptop or your system

[07:39] um you will be able to see the uh poll pop up on your uh screen and at that time you will have to fill your full name and um you know submit it in order to get the workshop recording and certificate.

[07:52] certificate. Okay, great. All right, so um we will Okay, great. All right, so um we will move on to the uh core of um the workshop. Let me just take like you know 2 three minutes to set context about um

[08:05] aentki and what we are here to talk about. I want to start with um this question here and I would really encourage everyone to answer in the chat and not just think about it. Um let us make this session a lot more

[08:18] interactive. Um so let us know what you think was one thing you did in this week um if you're honest didn't really need you. So um you know you could type out like some work task um that you felt could be um could be could have been

[08:33] handled without you. Maybe it was drafting some cold email um or sorting through like a bunch of requests to figure out which one was urgent or approving something that you had already probably discussed in meetings five

[08:46] times before. Um if there are any you know such incidents like that um let me know such incidents like that um let me know. Okay. Caroline saying type out notes from an online tutorial. Okay. Right. Yeah. That could be um very you

[09:00] know manual time consuming task. And there's someone saying researching. Okay. Right. Preparing audit documents. Okay. All right. Okay. Yes. Responding to emails,

[09:13] emails, MCP data structures, quality audit. Okay. Great. All right. So a lot of um you know different responses that have come in um and um you know what is interesting what is a pattern in this is

[09:26] that um what you're actually um you know it's almost never the task itself that needed you needed uh a human. It's usually like know decision making behind it. Um that needs a human which you know decisions like which email mattered

[09:42] whether to follow up or not what to say uh next. Um so that's the part actually uh next. Um so that's the part actually needs uh you right and um that's the difference um uh you know which is very key here because automation the kind

[09:56] we've had for years could um could do any task. Um it could send the email, it could move the file, it could fill the form. What it can

[10:09] know which one mattered and that's the shift we're talking about tonight. Um it's not just about doing the steps anymore. It's about deciding what to do next. And that one line is basically the whole difference between uh the tools

[10:21] you've been using that all of us are using um and what we what we will be building tonight um which is the agent. Um, of course, our expert is going to come in very soon and explain more about agent AI in a lot more detail. But let

[10:35] me just share uh this stat with you, $9.89 billion. The agent AI market is on track to hit 10 billion almost 10 billion this year

[10:47] alone and it's growing at 42% year-over-year through 2031. The growth of agentic AI is happening right now and it's happening at an extremely rapid it's happening at an extremely rapid pace. So naturally the question um from

[11:01] your end becomes okay so how do I learn agentic and how to get on top of what the job market demands right now or how do I actually um you know build an AI agent an agentic systems and a lot more right um well u this is what that is

[11:16] what this workshop and all that we offer after this workshop is all about so without any further ado let's uh look at the agenda of the session we'll start with about um half an on the core concepts. Uh some theoretical basics

[11:31] that will help you um you know make sense of what we're going to build today better. Um and um this is nothing heavy. We have a lot of examples. We have a lot of use cases um um and just enough information so that the practical build

[11:45] part makes more sense to you as we go. Then our expert is going to open uh with a quick walk through of the tools that we're going to use in this workshop which is architect and studio. and then we will start the build together live

[11:57] right from there. Uh we encourage each and every one of you to not just watch this uh build and then like you know try to recreate it from memory later on. Build alongside with us in real time so that any issues that you have any

[12:10] concerns any queries like we would be able to help you address it right uh then and there. And um yes, so we'll build the first half of the agent, pause for a proper 10-minute break, and then come right back and pick up the same

[12:23] build. This time connecting a knowledge base and refining it further. So, um you actually working very efficiently. Um and after the build parts are done and you have a live AI agent in your system, we will give you an exclusive preview of

[12:37] one of our most popular in demand programs, um the professional certificate in agentic AI and multi- aent systems. So um and based on uh the time that we have left, we will also do a showcase from maybe a handful of you

[12:50] um and um you know have our experts answer your question live um and then we will wrap it up. Okay. Right. Okay. So quick um you know uh intro on what we're going to build tonight. Um I think this is very very exciting and it's going to

[13:05] be a very useful uh uh agent as well. It we are going to build an second

[13:24] glitch. Uh, we're going to build an SDR outreach assistant. Um, so picture this. You give it um, you know, you going to give this AI agent a prospects company and it goes and researches that company on its own. uh then it pulls from your

[13:38] own knowledge base whatever you upload about your product or company to actually personalize the outreach instead of sounding very generic and at the end um it hands you a cold email plus two or three alternate opening

[13:50] lines as well to test. Um so if you ever started um you know stare at a blank old email not knowing how to open it um this agent is basically going to solve that and a lot more. So we are very excited to get um this started right. So um I I

[14:08] don't think I want to hold you uh from the um you know workshop any longer. So let's meet the two people uh who are going to bring the session together. Uh first up we have Anerut Narayan. He is the co-founder and chief growth officer

[14:22] you're hearing this name, keep it in mind for the uh that for that for the next few hours. Uh because everything you touch tonight runs on their platform and it's amazing. Uh he's also the founder of everything AI, an open

[14:36] through the noise and actually find the right AI tool for what they're trying to know it's extremely useful if you've ever felt lost in how many tools exist out there right now. And before AI became his full-time thing, Anerud built

[14:51] and ran two companies, Growth Spartan, a performance marketing agency, and Master Life, a life skills platform. He also wrote a book, Scale Smart, back in 2019, which has sold, um, thousands of copies. And, um, I think these days he spends

[15:05] part of his time advising early age startups, uh, AI startups particularly on exactly that, growing and scaling. Uh he is also a Colombia University graduate and um I think honestly outside of work his life reads almost like a

[15:19] wish list. He has lived and worked across South America, Southeast Asia, Africa, the US and Europe. He's completed a full Olympic triathlon. Uh he has summited um Stoke Cungry which sits at around 20,000 ft and he played

[15:33] basketball in college, university and uh at state level as well. So um I think through learning and growing and building something the hard way is um

[15:45] he's definitely not exaggerating because he is living it and um alongside him we he is living it and um alongside him we have Syash Man product manager at Lizer. Suyash comes from a technical background and has spent his career building AI

[15:59] products. Um, and before Lizo, he helped grow uh, Whoosh, a YC backed company from zero to multi-million dollar revenue and over 200 enterprise getting like, you know, enterprise

[16:12] customer to sign off on something, you know how big of a deal that is. He is also a former uh co-founder of Young Basket and he was in the top 5% of Next Leap's product management fellowship. And uh right now at Lizer he is building

[16:25] architect the exact tool where you're going to use tonight which lets you generate a full working agentic application from prompts. Um so the quite literally the person building the tool you're going to um you know build

[16:38] with. And quick word on lizer itself. I'm sure uh our experts will share more. Um but before I hand it over to them um let me introduce Liza. It was founded in uh 2023 and in a fairly short time has become one of the most serious names in

[16:54] enterprise AI agent infrastructure. Um with over million agents live in production and 30,000 plus developers building on it. Firms like Deoid and building on it. Firms like Deoid and KPMG Loose uh use uh Lizer uh to build

[17:06] AI systems for their own clients. Uh so tonight you're learning on the same platform running in production at some really huge companies. Anirut uh Suash welcome. It's amazing to have you and host you uh in this session. Over to you

[17:21] get us started? Um please. >> Yeah, I mean thank you for having us Ana. That was a fantastic introduction. I promise everyone we didn't ask Ana to give us that intro. She just picked that straight from LinkedIn and uh good to

[17:35] hear also. I learned a lot about Suy through this intro outside of this. So so thank you for that. If you want I can start sharing my screen and then uh should we do that? Okay, >> we can get that started. Yes.

[17:47] >> we can get that started. Yes. >> Cool. I'll share my screen. Um, >> Cool. I'll share my screen. Um, there we go.

[18:03] all, the fact that all of you are here on a Thursday, uh, you're winning in life. I would have not attended a webinar on a Thursday. not attended a webinar on a Thursday. So, so kudos to you. And um I think the

[18:15] way we want to do this is it'll seem theoretical. It'll seem technical the first 30 40 minutes. It might feel like, man, why are you overwhelming me with man, why are you overwhelming me with these concepts, but I guarantee you it

[18:27] gets easier as as we move in this webinar. Uh so I'll cover the the harder technical stuff or the semi-technical stuff. And I feel like most of you would know a lot of these concepts. So, but the way we're going to try to do this is

[18:41] I will show you what is an agent. I will show you parts of studio. I will also cover maybe some of the use cases that we've actually shown as prototypes. So, we've actually taken them into production. And then Suyash will kind of

[18:54] say, okay, if all this complexity could be abstracted and made easier for you so that you can actually build an agent by speaking and a win would be you're able to take it live today, that would be huge, right? And we're hoping like we

[19:07] could kind of announce a top three at the end of the day in terms of here are your top three uh folks who have built out agents. We could give them maybe some gift cards and stuff. So So that's what we're going to do today. Um so

[19:19] before we get into like what is what is an agent? What is like what is what is an agent? What is what do all of you think is an agent? what do all of you think is an agent? Let's see what's in the chat.

[19:32] Feel free. Brashant joy. What is all of you? What is an agent according to you?

[19:47] its own without intervention. Okay, great. The one who does the instruction given by us. That's pretty good. A nonhuman system that achieves a task.

[19:59] Um, someone said no idea. Great. We will cover some of that. Uh, so the way I think about I think before we get into what is an agent, I want to give you a larger perspective what's happening in

[20:12] the AI world. Right? So today obviously there's a big need for energy. All the energy is mostly coming from uh coal, petroleum, some of it's obviously renewable sources. There's a big fight for getting energy, right? So that's

[20:25] actually maybe at the bottom layer here. Now that energy and those minerals essentially are being converted into chips which Nvidia has thrived in making designs for these chips and TSMC I would say obviously in Taiwan is manufactured

[20:41] 90% of those chips but those chips have become an integral part of training a lot of these models okay and so you have Nvidia at the most bottom layer then you

[20:53] have these GPU providers so what is the GPU providers so AWS Azure Google obviously you have Meta what they have done is they have actually bought these chips from Nvidia for billions of dollars. Why? Because they want to rent

[21:06] out these chips for compute. So people want to use especially LLM or like AI models or run these models will run them on these GPUs. And the cool thing about

[21:19] tasks at scale. So that's why you've been able to do you need GPUs for a lot of this training. So they provide these GPUs and then you have these AI all of you most of you I'm sure are using claude there's chat GBD mist is

[21:36] based out of France a lot of them have decided to take different directions in which they built foundational models open obviously like if you think today I feel like it's become more of a B2C product anthropic is becoming great for

[21:49] businesses mist hitting the sovereign AI story just to give you context mist France. They're also moving into a physical hardware like product. Hugging face is like hey you know what we'll create open source models and we'll show

[22:04] you a lot of these open source models that you can use. So uh just to give you so you don't get to see what's happening under the hood. Open source obviously hugging face you get to see how these models have been built. Okay. So this

[22:18] in and says I want to either build these models I want to help fine-tune these models. So you do fine-tuning because maybe you want these models to be specific for a certain industry. You don't need let's say a trillion

[22:32] parameter trillion token or trillion parameter model. What if I need a 7 billion parameter model which is just focused for finance. So you have a lot of these companies that help on the finetuning, they help on the training.

[22:44] That's essentially layer one. Layer two is data. So companies that are thriving today are the ones that own the data. Take someone like a Salesforce. Take system of records. So everyone's kind of building AI on top of that application.

[23:00] you have data? You have something called a structured format and you have an unstructured format. Structured format is it's in a form of a database. Let's say so you have MySQL, you have let's say Postgress, you have a lot of these

[23:14] uh databases and there are companies like data bricks, upstine cone, they've built companies on making sure all of this data can be easily accessed within a company. Then you have data prep-processing where a lot of the data

[23:27] for LLMs have to be converted let's say into a vector format. So companies thrive in that. And then there's unstructured data which is let's say a PDF is an unstructured data. Image is also an unstructured. All of those

[23:39] companies are helping convert some of that unstructured data into structured data using algorithms. So that's layer two. Layer three is deployment. Deployment is where we are essentially building agents. So I'll tell you what

[23:52] an agent is. But deploy lizer is where comes in the agent tool framework space. What is a framework at the end of the day? I don't know if you've heard of let's say this Django framework which is used for build applications. You have

[24:06] Unreal Engine and gaming which is a gaming framework. You have Ruby on Rails which is a coding framework. So your framework the way to think about it is it's like a kitchen. It provides all the tools. It provides the stuff. It

[24:18] provides all the ingredients for you to build something. So agent tool framework is where Lizer is and we sit in this category along with lang chain autog fix. Now there like 200 different builders or agent framework layers that

[24:32] are come to the market. What makes Liza different is we are low code. We are great for enterprises because we are we own the entire stack which I'll tell you what I what we mean by that. Uh so and each of them start specializing in

[24:45] specific things. So deployment means you're taking it live and lizer in some sense make sure all of this is available within our platform. Deployment is you're taking it into production which is actually into an environment where

[24:57] customers can interact with it. Orchestration means I might have different models I need to or I might have multiple agents. Martian may maybe acts like a router which is oh now if there's a frontier model coming in or is

[25:10] there's an input that's coming in how do I route that task into specific models prompt management that's what started with and langu but there are specific integrated. And then finally, observability, which is you want to know

[25:23] what's happening with these agents when they go live. Okay, so a lot of you said agents are something that does autonomously. They can do a specific task. I'll get to that, but at the end of the day, the way to think about LLMs

[25:36] is it's been trained on almost all the world's data. You give it a prompt, it converts it into a vector format, which is basically, I don't know if you remember back in school, it was like that matrix which had like these three

[25:49] points. They all have weights. They all have probabilities and it tells you what the probability. So for example, if I say my cat is dash the next word is all probabilities of what the it's a predictor in some sense. So what it

[26:05] helps you do is LMS basically help you decide what the predictive text could be. And in 2017 there was a big breakthrough which was earlier you could only predict uh words then it was sentences and then with generative AI

[26:19] you moved to paragraphs. So there was a breakthrough in the model which is basically the transform model and now you've moved to that application of that has moved into I'm building LLMs not for just for text generation but for video

[26:33] generation for image generation you name it all forms of voice generation for all all forms of products. Okay let me see if there are any questions if there are any questions uh

[26:51] let's talk about agent. Okay, the way to think about an agent at the end of the day is before agents was let's say robotic before agents was let's say robotic process automation or you creating very

[27:04] hardcoded rules. So, for example, if I wanted to research about a specific player, I would have been like, okay, I want to research about simply learn. I'd be like, hey, go to first go to Google, fetch this information. If I don't find

[27:17] it on Google, go to crunchbase. If I don't find on crunchbace, go to uh LinkedIn, right? I would provide a lot of hardcoded journeys on get this in on What agents have done is you just need

[27:32] to give it a single input. Hey, get me information about simply learn and you connect it to all these tools. Connect it to a crunch base. Connect it to a it to a crunch base. Connect it to a LinkedIn. Okay. Connect it to uh

[27:47] what else? uh let's say a Google search API. Once you do that, you also connect API. Once you do that, you also connect it to an LLM. Once you do, it automatically decides where to get this information, prioritize it, and give it

[28:01] to you in a format that you want. Now, what you're seeing in this diagram is what you're seeing in this diagram is basically LLMs are like this beautiful abundant water and agents are you're adding shape to the water. So when you

[28:15] say shape, it is you need to give it shape in terms of you want to make sure it has short-term memory. You want to make sure it's long-term memory. So when you'll be like it has a session, it's able to understand what's happening

[28:29] here, but maybe like if I come back and I go to a different session, it doesn't remember it. Tool calling is you're connecting to different tools which I just told you. Okay. Agent examples is maybe you're giving it context on what

[28:43] the output should be. Hallucination check is you want to make sure like hey the output that is coming out is reliable. So all of these factors we call them modules. Okay there are four things to think about. You add data, you

[28:56] add an LLM, you add tools and then your modules. The four things come together for you to actually build an agent. Okay. So the data as I remember showed you could be structured unstructured data. modules are your hey I want to

[29:12] and I'll show it to you in terms of how that looks and then finally you give it specific u input in terms of what the what the example should be and then from there you connect to a tool okay

[29:29] maybe just show you straight off the bat what is an agent so this is what a platform looks like by the way so when I say I can go to studio.licly.ai AI and if you're front of a laptop absolutely go there I have a home I have agent

[29:43] registry I have create agent I have voice now let's say I go to create agent voice now let's say I go to create agent and I say agent okay now whenever you're thinking of creating an agent it needs a role so you think of it as you're

[29:56] telling this agent hey by the way you are an you are an UIUX expert you are a customer support expert you are an agent that does customer support. So that's

[30:10] the role you're giving it. Then you're giving it a goal. Hey, I want you to answer specific customer questions. Then you're giving instructions. Hey, you know what? Instructions could be around even specific tasks. So you could add

[30:24] specific like FAQs. You could add system level tasks instructions. You could do show you is you don't have to do any of this. You should just be able to say, "Dude, I want to build a AISDR agent." It'll automatically pick RO goal

[30:38] instructions, but I'm showing you what's under the hood. Okay. So, you click on roll. You click on goal. Maybe I shall go back. I'll show you one of the go back. I'll show you one of the specific agents. How about that? Um, so

[30:51] I go to agent. It's taking some time to load. 102 agents. So, let's come here, right? Okay. Let's say a partner data agent. Here's the role that says you're a partner data retrieval agent, right?

[31:05] company record. structured company records as I mentioned was like a database right goal is accurately retrieve this information instructions is you can be specific your knowledge base contains so knowledge base is what

[31:19] where you pulling data from contains information some of this it says your response must follow the JSON schema don't don't worry about that this is basically saying hey I want to make sure it's in a specific format and always

[31:32] follow this format when you're giving me the output okay now you can do all of this now this is part of the agent. What I tell you first, you're giving it intelligence. So you can pick multiple

[31:46] models. Now what we talk about you have multiple models. You have open AI, right? OpenAI has come up with a new I think GPD 5.2 which which we have here. So what's the advantage of a nano model like maybe it's great for like specific

[32:00] tasks. You want faster output, right? Then you have O4. I think it was a model. So you have a lot of these models. So we are giving you a platform where you can pick and choose a model. Okay. If you're a developer, you might

[32:14] be like, you know what, I prefer anthropic. I want to use Opus 5 which is the latest model or I might want to use Haiku which is faster but quicker answers but it's like the task is not a complicated task. Right now we you can

[32:28] also pick models from bedrock. So Amazon what they have tried to do is they want to continue pushing their revenue. So they have built their own models and they've also built partnership with a lot of existing players. So Mistl which

[32:41] is based out of not Europe their models exist here your GPT someone GPT models llama models which is from meta nova deepseek some open quen which is Alibaba

[32:53] if I'm not mistaken all these models are available GLM if I'm not mistaken actually is some of Gemini sorry Gemma is some of Gemini models okay so all of this is available here now how do I know which model to pick based on the agent

[33:06] that I'm building now if I come here suggest model actually let's do this if I suggest model you might want some models which is based on speed you might want some models which is based on cost right so

[33:20] and it depends on the task right some maybe if you're doing a long form like hey give me research you're trying to do like a deep research speed may be less important and you may be willing to pay pay some on the cost but if you want

[33:33] quick answers for a customer speed may be really important so this is a second part of your agent. Third one is knowledge. So when I say knowledge right, let me show you. When I click on knowledge base,

[33:48] knowledge basically is I have knowledge can come in the form of um structured data and unstructured. So I structured data and unstructured. So I can upload Excel files, I can upload

[34:02] uh CSV, I can upload. So all of that. When I click on new, if I show knowledge When I click on new, if I show knowledge base, yeah, I can name the the file, it format and then I can create the knowledge base. Okay, so all of that you

[34:19] can do from here. You can also do it directly from studio, right? So if I come here uh uh knowledge, yeah, I can add once something, I can add it directly from here. Okay. Then I have tools. Tools is

[34:36] where you truly decide what you want to connect to to automate tasks. So if I look at tools here, let's say if I come here, monitoring connections, I look at here, monitoring connections, I look at tools. Tools is what you wanted to

[34:49] workflows. You want to connect to a Gmail. So on the AI SDR side today, you will actually connect to a Gmail, but you might want to connect to Slack. So really good at when you're building agents is what does the overarching

[35:02] architecture look like? Where do I want to pull this information from to do let's say a specific task? You might also have to think about if there are actually have an agent for each of those tasks? Okay, but here you're essentially

[35:18] adding a tool and then finally you have something called skills. So where it's moving today is earlier which is maybe even like a couple of months back you had you would have multiple agents you can have a managerial agent and then you

[35:30] have one agent's output going to an input of another agent. Skills is you can basically convert any like let's say you have a branding you're giving it can convert it into a skill and add it as a skill towards the agent and then

[35:44] you can run automations okay and then the features is what the modules are do you want to have memory should it be able to be should it be should it be able to be should it be safe and responsible all of that is

[35:57] safe and responsible all of that is literally being defined here about at a high Uh before we jump into the specifics of

[36:09] like me giving you examples of different forms of agents any questions so far

[36:21] how to join this is not free uh our studio is free to join so you can definitely use it uh lizer is a multi- aent platform yes you can build multiple agents agents software system. Okay, cool. Are we good? Okay, I would, by the

[36:36] minutes, if you're doing this over your phone, definitely switch to a desktop because I'm going to start showing you something interesting as well. Okay, so I told you what is an agent. All of this stuff that I showed you is finally I

[36:50] showed you as part of the studio and agents are the connective tissue. So what I tell you now you have you you the future in a lot of ways is you end up building agents for each part or each function. Now you will build an AISDR.

[37:05] Now let's say somebody else build a built a customer service agent. What if they could speak to each other? So you have that option here as well right? So one if let's say if I go here generate improve

[37:18] improve update places share uh uh connect get version history publish clone there is a way once I

[37:31] deploy this agent as well I can as an API share this agent with somebody else okay so that's how you start building multiple agents and you start connecting with agents across a period of time Okay. U now I think Suyash will show you

[37:46] that as well. Now what when it says how it started right earlier it was a lot around code. You have to define what and by the way just to give you context in terms of what temperature and top P

[38:00] in terms of what temperature and top P is right if I come here it's finally comes down to probability. So if I type let's say online marketing agents right this entire list right these are all probabilities that

[38:14] have been built based on what people are saying right let me keep it online saying right let me keep it online marketing a marketing for huh when I do this these are all probabilities when I say temperature I have the option to

[38:26] pick a temperature between 0 and one right which is what is here right when I say temperature is seven you're basically saying here hey

[38:39] consider 70% of the probability. So for example, is a 25% probability. This is a 30% let's say 15% probability. The probability reduces as you go further down of the next predictive word. If I

[38:54] said temperature needs to be 02, it'll only consider let's say the probability of that word. So you make it makes the output more deterministic. If I made it Z, if I made it 0.1, the output will always be online marketing

[39:07] for small businesses. But when I say temperature on a scale of 0 to 1, if I said temperature is 0, let's say it's one, it considers the entire list of all the options of probabilities that can show up. Whereas

[39:21] if I made it, let's say 85, it'll maybe cut off the last two and it'll only consider this sample set. So for you, the way to keep it simple, the way to think about it is you don't have to first of all think about top and

[39:36] probabilities. Okay. But in this scenario, if it is, if you want the results to be very accurate, you want to reduce temperature. sample set, you want to reduce temperature. But if you want it to be

[39:51] trying to build images, you want to get as many as many variations on videos, that's when you can increase temperature and talk. Okay. So how it started it's

[40:05] started with this which is hey you give it model temperature top messages give it a role which you saw there on the low code version and then it's moved to man add all these modules around it okay now what is the difference between agentic

[40:19] AI and AI agents what you just saw was an AI agent with all these features agentic AI system is where all of these agents are connected to each other and will build today most probably an agent AI system persistent memory is it will

[40:36] remember even when you come back inter agent communication agents able to speak to each other okay what I showed you is a single agent system you have an input goes into the agent it accesses a knowledge base accesses tools processes

[40:49] it goes to the output what you will build today most probably is a multi- aent system what is a multi- aent system it'll not look this complicated but if I had if I was building an agent where I needed information coming in from chat,

[41:04] I need information coming in from phone. All of that information if you wanted, I for each. So chat could go into a conversational chat agent. You can be like, "Hey, maybe this information is in text. This information is in voice.

[41:18] Emails may be a combination of multiple formats. It's text plus images." So you separately, right? Slack could be message handling. So you could define four of these could come together and be like hey maybe I need a master agent to

[41:33] look at that information resolve where do I take that information who do I pass do I take that information who do I pass it on to so if it's answered go down if it's unanswered go up and then this is basically another place on where the

[41:45] data is coming I might have historical tickets which is your knowledge base I might have a third party API system which is like I'm collecting data from things to outside of my system right knowledge search I told you is

[41:59] updated and then deep learning agent you don't have to think about that too much but basically the training sets that you've added let's say from hugging face or any other place to add more context to this so what you will do today most

[42:12] probably is build something like this less complicated obviously and you have multiple formats of multi- aents what you see is a single agent but you could have an agent where hey so in this network is if I have four agents and a

[42:26] to each other. A manager agent could speak to this. And I'll give you an example with an AISDR, right? With an AISDR, let's say today you'll build like, hey, I want to do research on Suyash. So now when I do research on

[42:42] Suyash, you might want to hit different tools. Then that information has to be passed on to another agent, which is it's going to look at that information and compose an email. Then let's say compose an email. that compose email

[42:56] bring in before which is before sending it and then it sends the agent sends the information sorry now let's say somebody responds now somebody responding you want an agent to look at the response look at that information and then

[43:10] provide a separate response so now each of those agents could speak to each other right or you could have them sequential so it'll come down to how you actually build some of these things out supervisor very and some of them

[43:23] actually if you surp you won't be surprised. A lot of them actually mimic how companies are built. If you have a company with a manager with each of them doing individual things, you'll most probably have like one person that just

[43:37] agent. One person that just does image creation, sorry, uh ads might just have ads agent, right? So there is some similarity in how organizations are

[43:49] built as well. So you have different multi- aent patterns. You I told you about managerial self-improving. What this might mean is you might have another layer here which is if I have an input I have an agent that comes in

[44:02] tools are connected to it. Now it goes to agent two. Now let's say for if it fails for some reason it can go back to agent one. So that's the roll back. You might want to in some sense have so this this you have the advantage of it being

[44:15] cyclical. You might have one set of agents which are just connected to in this this almost forms like a sube ecosystem. Agent three is broken down requires a lot more. So you have to really start understanding workflows of

[44:28] an organization. I can tell you prompt engineering is is a skill that was super hot last year. Right now the hot skill is are you really good at orchestrating the or understanding the workflow in

[44:43] converting into agents. That's what Aifds which are for deployed engineers are getting paid for. Where does human the loop come in? How do you add tools to it? All of that is the is where actually where the money is. Okay, we

[44:58] have deterministic versus agentic workflows as well. So determine not all workflows need to be agentic. What's the problem with agentic. What's the problem with agentic? Some of the results can be like

[45:12] when you want highly accurate output. Let's say in a financial firm, you want zero hallucination. what maybe you don't bring agent into that workflow but you bring an agent into the reasoning side of it what does that mean if I wanted to

[45:25] hit a database pull that information out and I'm like hey just pull let's say population data of China India all of that or let's say just all Asian countries break it down by multiple formats that could be agentic but what

[45:38] formats that could be agentic but what if I said hey add the two and like let's say like hey give me just some cluster information as a sub subset you could have that as an automated workflow where you're using not agents but you're using

[45:51] deterministic workflows and then you can add an agent which is on the reasoning layer. So some a lot of the implementation that we do today you'll termination workflows and agent workflows.

[46:05] Okay, autonomous agents is you would take it one step further when there's no human in the loop. you want to do a specific task, it keeps doing these tasks and then it gives you an output when you are

[46:17] then it gives you an output when you are like and it could be right from I'm looking to shop. Hey, buy me show me a product which is I'm looking to buy shoe size 11. I'm preparing for high rocks. Give me these shoes.

[46:31] Autonomously, the agent might be like, "Hey, I'm going to look for all this "Hey, I'm going to look for all this information. Uh maybe even make sure it is on the checkout." and then all you have to do is review it. Right? So

[46:43] have to do is review it. Right? So you're giving it autonomy in a specific workflow or a loop. One of the things that we were evaluating right now is be able to let's say I've decided to create an event on Luma.

[46:57] I should be able to say hey I want to find people for this event. So do an outbound is done it'll go hit let's say Apollo it'll hit Clay. It'll find people from LinkedIn. Let's say it starts sending emails and it's you've already

[47:11] doing that, let's say people RSVP, you can have autonomy on RSVP automatically. You can also have autonomy on, hey, by the way, once it automatically signed up e-commerce tool, let's say like a Shopify, where they're saying, dude, I

[47:26] can automatically check out and buy merch and send it to them. So that's the level of autonomy you can give. But today the way the way it's set like I want to decide by human in the loops I so there are specific task there

[47:41] is there's autonomy but there's like mostly human in the loop in most of mostly human in the loop in most of these workflows that we build.

[47:55] you maybe u some examples maybe that could be interesting. How are your time by the way? way? Anybody?

[48:11] >> Okay, great. Now, let's talk about applications. I think this is the exciting part. So, now what are people doing? People are like, "Okay, I've built these agents out. Let me go one step further and build out the workflows

[48:24] things that we did is we created these blueprints or we created these pre-built agents. So we have pre-built agents for banking. We have pre-built for insurance. We built it for marketing, sales, HR, customer support where we're

[48:37] saying we will capture 60% of the logic and then we'll customize for you and take it live. Okay, what does that mean? Now let's actually take some of the examples. This is an agent we built internally within Lizo. We call it

[48:49] Scott. And what does the agent do? It's built for the marketing team. I should be able to create content. But what makes it agentic is when the agent is telling me what to do. I think there's a big difference also the way you want to

[49:01] start thinking about this. SAS can also pull this information out. Okay. But when when it feels intelligent, when it starts telling you part. So I'll give you context right now. Let's say we're trying to build

[49:15] blogs. It'll show me, hey, here are all the blogs we've published. Here is my they've added some dashboard layer to it. Now I say hey dude tell me pull all the information associated with us. So it says okay Palunteer reports 85%

[49:29] growth. Glean is doing this. So I've put a comp set of comparator names and ancillary company names. So I can look at all these articles. Then I have community tell me what's happening within Discord.

[49:41] Tell me what's happening within Reddit. So I pull information. Then I have Lizo. What's happening within Slack? what's happening with HubSpot. I'm using all these signals and then from there I'm saying give me content ideas. It's

[49:56] agentic. What's the difference? It's saying okay volume is this. So here's comparative content you can create. Here's SEO focused content. Here's happening. And then let's say you've uploaded. Now I can go to SEO and I can

[50:11] add this to my calendar, right? And then I can actually generate And now this content will go sit on WordPress. So what what are we doing? We

[50:23] are making sure content can be created, content can be synthesized and then placed on WordPress. And there's a human in loop within WordPress. Okay? And we

[50:35] have done this for all the other ones. So if I had to show you while this is being built, right? Let me show you another one. Yeah. Now what is that? This is another

[50:49] knowledge base. We uploaded all our case studies. We uploaded text. We uploaded studies. We uploaded text. We uploaded URLs. All of this is here. Okay. Social library is we've also started start start hitting hey where is this? This is

[51:02] start hitting hey where is this? This is maybe not as real time but like we're able to pull in even social data from LinkedIn or Twitter and show up here. Okay. Then we have something called LLM wiki which is what is the data set and

[51:16] what are the tools waiting what is the agents that we have internally right and then we have for everything we have for email if you want to create an ABM outbound you can do that if you want to create ads you can do that you want

[51:29] to create LinkedIn content you can do that very similar we have that for all the platforms okay you can create content um ABM outbound is you can pick an agent,

[51:43] you can pick an agent, you can pick uh a hub. So you can connect to say okay I want to send specific information to an account and then it can actually context. So we have connected let's say HubSpot we connected to a noteaker right

[51:58] so all that information you can pull out so this is let's say Scott now if I want to go one step further let me show you HR okay HR agentic OS is what if I want HR okay HR agentic OS is what if I want to automate my entire HR work cycle

[52:13] right if you look at HR it is what I have uh employee onboarding I have resume filtering right I have recruitment I have offboarding And then I have HR productivity. If I had to do something

[52:27] as recruitment as talent pipeline, I I can build an agent goes to be like hit workday, hit greenhouse, hit Gmail, hit Google calendar. By the way, LinkedIn makes it hard for you to connect to the API and pull that information out. But

[52:40] it'll can also show you who who are my today's interviews. Now I can be like, hey, which requisition have been open and why? I can ask specific questions. agent will look at all this information and pull pull that out saying okay here

[52:54] is why it's taking so much time it'll hit all the tools and give you an answer hit all the tools and give you an answer okay here's the breakdown here is the what's in critical what's in red what's in green right then I can do candidate

[53:09] sourcing candidate sourcing hey search across LinkedIn GitHub and Indeed search let's say specific agents with this it again hits an agent pulls that again hits an agent pulls that information out So the this agentic OS

[53:22] has multiple agents sitting on the back end. You technically have an agent for each of them connected to each other with tools. So this is how it starts getting complicated. I mean from an from an email from an

[53:36] infrastructure standpoint but this is what customers are asking for. This is potentially 100k product right and they might be like hey I want to make sure my data is private. Can you deploy on my virtual private cloud? Can you make sure

[53:49] my data is maybe onrem? They might come and say listen uh Ani and Suyash I want you to not use I'm in the just the Google ecosystem. I want you to use only Gemini models. So that's where some of the customization comes in with AISD

[54:05] that you're going to build. Most people come to us with the customization is okay this is cool. Can you also send it on WhatsApp and SMS and Gmail? Let's say can you do multiple languages? Can you hit multiple tools? I want to in in the

[54:20] in the HR scenario, you'll realize they're like, hey, no is really expensive. I want to hit another tool where I can pull this information. where I can pull this information. Okay, so you have let's call this

[54:32] Okay, so you have let's call this agentic OS. I have this for HR. Let me show you more. We have something for CFO, which is you want to automate an entire CFO life cycle,

[54:44] to see what's happening with financial reconciliation. I can run a reconciliation. Um and again like what makes it agentic is you are able to come here. You're either able to ask specific questions.

[54:57] telling me hey what I should be doing next. Okay. So that's become the expectation. Now what you will be able to build today which is I'm pretty excited about you will build a front end like this

[55:10] without knowing how to code. It'll have sample data. It'll capture a lot of this nuance where your strength will be is man do I really understand the work. Okay. So I can build a CFO office agent OS which will be a monthly close

[55:25] regulatory capital. Now remember if I'm someone who's in marketing I don't understand CFO you can claw your way through and capture some of this nuance stuff but the person that's winning today is like they have really captured

[55:41] the nuance and they've gone vertically integrated Harvey for example is 11 billion company what have they done you had X lawyers coming together to build out Harvey $1 billion there are 700 people out of 700 350 are lawyers who

[55:54] act like product managers which are all capturing the nuance of agent. So agents if you're good at knowing what the limitations of an agent and what can it can do and you add your layer of your domain expertise that's

[56:09] layer of your domain expertise that's where you win. Okay. So I have my HR OS I have my CFO OS let me show you we have something we did for HFS research. So one of the things in the use case with HFS research was let me see if I have

[56:37] is they have written let's say thousands of articles okay and if you look at a lot of the search that was done back in the day was semantic search which is what I have Google Drive. I type keyword, it

[56:50] files and gives it to me. But they were like, hey, listen, can you look at can you make it agentic and what does agentic mean? I should be able to look at all the articles and tell me, hey, what is the trend that's happening with

[57:04] SAS based on articles? What's hap where is the mining industry headed? Those things you should be able to build using their website. So you go to hfsrearch.com. We built this like what

[57:20] specific questions and then obviously you can read each of these articles. So very perplexity style but giving you agentic output from a place of

[57:32] summarization and reason. Okay. So this is another use case that we seeing for uh research. Let's look at maybe one more which is the CBC agent. So Axenture

[57:44] is one of our customers just to give you context. Lizer is around 200 people. We are we do a lot of agent building for companies. Accenture is our customer. They're also our investor. US government, Axenture,

[58:00] government, Axenture, um KPMG, then you have Crown Castle, WDW, they're all our customers. Okay. And what we have been able to do really well is we've been able to take complex use cases because taking agents into

[58:16] production or taking them live is not just a AI problem or a data problem. It's actually also a people problem. How do you make sure people truly understand

[58:28] how to capture the nuance? Do you have a champion internally? Do you know where all the data sits? So what we have tried to do is we try to own everything end to end agent building agent hardening which is basically

[58:41] agent hardening which is basically refining this agent agent deployment and then agent governance once it's live okay and it's a platform plus people model equal to productionization so and the way we showcase our capabilities

[58:53] actually building on a lot of these use cases so spotlight is an accenture product okay and we we launched this partnership with Axenture which is over the next three years can we do 10 million error with Axenture can we sell

[59:06] this to other venture capital arms especially on the corporate side can we sell it to Barclay's ventures can we sell this to let's say Qualcomm ventures okay list all the climate tech startups so

[59:21] maybe I'll see if there is huh I come and ask hey show me all the climate tech startups so it says I have identified 60 plus climate tech startups now what's the cool thing about each of this Okay, I can view the evaluation based on

[59:35] Axenture has given us a 28 point framework on if we had to invest this is how we would rate it. So it says here Aenture enterprise readiness 5.2 two founder and cap table 6.5 market there it's

[59:50] promising but early technology you're too early products early right they also say strengths are here so it gives you a summary now you also have something called bessemer inspired s sov

[1:00:09] score on a scale of 100 okay and then you are able to now the next step with this Not just on the analysis automate founder interviews. You can also build memos. So if I say here read

[1:00:24] assessment. I can download this report as a PDF. as a PDF. I can also send this to my team, right? the founder interview. Now the founder interview is if I want to do a new

[1:00:38] interview is if I want to do a new interview let's say Suyash is a YC founder and I have gotten his information from Crunchb and all these places and by the way accenture spotlight hits multiple places once it

[1:00:50] does I might be like dude I need the this information is dated I need more information from Suyas so I might be like okay ready let's ask specific information from yeah let's let's ask specific

[1:01:06] information from Suash. So I can basically set up interview questions and then send this as a link to Suash and then Suyash will receive a voice agent that he can in some sense go back and forth with.

[1:01:19] Okay, so we've built a lot of these agents. You have Spotlight News. It shows you here's all the stuff that's happening around the world. That's the that's one more use case of these agents running in the back end. To give you

[1:01:32] context with Axenture, if I'm not mistaken, let me see if I can show you the uh look at this. I have 200 agents running in the back end. I've hit 40,000 data points. And it says contextual signals map. Basically, once you map all

[1:01:47] these data sets uh data points, you get 5 billion contextual signals. So, it's like it's a pretty complex agitic system on the back end. Okay. So I've done I

[1:01:59] showed you HR I showed you CFO I showed you HFS this and we built a lot of these pro prototypes I have something for procurement

[1:02:12] where it's like can I manage different suppliers okay I won't go into the specifics of this but let's say if we have something for all

[1:02:30] we built for different customers. Okay. And what we are seeing in the market right now is sales, marketing, HR, customer support, really hot but very crowded. IT ops, procurement, CFO is the second one that's picking up and this is

[1:02:45] how you're seeing a lot of these use cases explode. Now where you have to get good at obviously is understanding agentic workflows. If you're a developer or if you're someone who's has an engineering bend and likes to build

[1:02:57] stuff, what you have to get really good is capturing business logic and building out beautiful user experiences for people. Building products has become really easy. What you if you can do is capture nuance, that's where you win.

[1:03:11] market, someone who is good at sales, someone who's good at marketing, someone who's good at product, what you have to get extremely good at is what is big picture? How does all of this fit? How do I sell this? How do I get

[1:03:24] distribution? Distribution's become hard. If you look at the entire where the world is headed today, take videos. I can build videos with Gemini. I can build videos with Hicksfield. I can do this with uh what

[1:03:39] is the uh with Hen. Video generations become easy. What has become hard? Video editing, right? How do I capture the last man? Similarly with agent agent building has become easy what's become hard deployment right so last mile is

[1:03:55] where the hard part is and that's that's where you have to figure out how do and that's where the most values also I give a good example of like I don't know if a good example of like I don't know if how many you watch Naros Naros uh on

[1:04:08] Netflix right they said when they were selling cocaine cocaine used to be $8 okay a gram but by the time it came to the US it was $100 a gram And it's why because there was most risk in distribution moving from Colombia to

[1:04:23] Miami. So the massive value generation was there. Lizer operates in that last mile massive value generation. But something for you guys to do is we we own the entire distribution. And that's why you

[1:04:36] have to also start thinking about if you build a brand that's a mode. Okay. Um cool. So that's where agents are headed. These are just some more under the hood with Nvidia. We

[1:04:50] did a use case which is which very similar to your AISDR play that you might do today. Inbound leads coming in automatically evaluate where it's coming in from. I should I should automatically add a lead score to it. Send automate

[1:05:04] the emails being sent and then I should be able to have conversations in AI and anything it's unable to answer send it for training. Okay, so Mato was one use

[1:05:16] case that we had explored a while back where what they want to do is automate spent so much time on performance reviews. What if I had a ninebox frameworks on how we automate based on all the information that's captured

[1:05:28] about a user? We should automatically put them in a box. Give them a plan on what their coaching plan in terms of getting uh getting better as well as like maybe even decide from their last mile on what their variable should be.

[1:05:42] Okay. Autonomous AI recruiter. I showed you the HR one marketing content generator. Today, this was actually surprisingly quite relevant. Uh it's still very relevant today. But because content generation has become so easy

[1:05:56] with Claude. Um today it's become about do you capture the nuance of can you do long form content at scale? Okay. So that's where we were thriving with marketing content generation with Hitachi. Okay.

[1:06:09] What I showed you was Accenture CVC, right? All this data is where we're capturing information from, which is what Gartner, G2, Pitchbook, Apollo, Crunchbace, right? Then it's like, oh, for this specific company, I'm getting

[1:06:23] information from newsletter, LinkedIn, website, search API, founder review, Google news, right? Then internet search, go after perplexity, Google newsletters, and then customers uh custom list is also being added. And

[1:06:37] graph. Now what's the advantage of a knowledge graph? All that information imagine whether a comp person leaves the organization or not that information always stays within the organization. We were talking to a legal company. They're

[1:06:51] like man half our conversations on WhatsApp and when somebody leaves there was a way where all this information can be captured week on week out and then there's a knowledge graph that's being built. There's a reason why

[1:07:07] what you're seeing today Salesforce has so much data with them. It's very hard to rip and replace Salesforce. You have this data that sits on Salesforce and then from there it's able to in some people are building agency systems on

[1:07:20] top of Salesforce. Okay. Last thing is Shiva Claw. Okay. So this is we built a fundraising agent. You might find this exciting as well.

[1:07:32] What we did is we raising a series B. So we're pretty close to closing our round we're pretty close to closing our round and what we did is what if I mean if fundraising people spend a lot of time giving information talking to analysts

[1:07:46] let's front that information instead of you spending four weeks just doing this fund raise what if we said okay you know what why don't people can have questions so you can come here you can be like hey what does Lizo do you can ask specific

[1:08:00] questions right you want to ask specific questions about our tech which is gent protocol you can you want to ask specific questions around Shiva Jimmy me you can all of this you can do okay you want to ask specific question on

[1:08:13] want to register interest you can register interest okay and remember as I told you people it's moving from multi- aent systems into single agent with multiple skills so here's a claw agent it has a soul which is I'm adding

[1:08:29] information like how remember you had goal role all of that that you can add as part of the soul rules is hey what does what I want you to do instructions knowledge is where it's pulling this information from skills is here are

[1:08:41] different skills it taps into to build things out Right? So you can register ask all these questions and then it also builds an investment memo.

[1:08:53] Right? So we ended up having close to 131 investors fill this out and we had 17 investors almost go to the last round. So we were covered on Techrunch, we were covered on Bloomberg, we covered on inc.com, right? Uh so we see AI agent

[1:09:11] startup that lets his agent run $100 million fund raise, right? So this is you're able to in we did this actually surprisingly with agent Sam for a series A as well. Okay. When I say agent Sam

[1:09:26] uh it was something like this. Okay. Agent Sam maybe lies. Right. Let me see. Let me see if I have a video. Yeah, this could be an interesting video. Where's the video? Huh?

[1:09:43] right? The full stack. >> So, what we did is similar. We built a landing page. We had an agent and an agent here is basically like >> we're launching our serious a fundraising. But instead of doing this

[1:09:58] the traditional way, we actually built an agent on Lizer. Meet agent Sam. >> This is Sam here and it's nice to meet you. Agent Sam will interview me asking all the questions that you as an investor may have about Lizer.

[1:10:12] >> Thanks for the kind introduction SA. So let's start with some basic questions. What is Lizer and why did you decide to build Lizer? >> I mean if you can see the similarity this is a combination of Mark Zuckerberg

[1:10:24] and Sam Alman. So it looks like Sam Alman the busker. But what we essentially did is uh we built a video like this. Shaz asked questions and then it went to a landing page. Okay. So this is uh this is essentially the the

[1:10:39] to get really good at. What will you build tomorrow? Okay. Um so I will leave it at that today. I think if I had to just do a quick TLDDR

[1:10:52] um actually I'll leave you with one more thing and then we'll get to this. Now you might be like dude why can't I use claude? Why can't I use cla co-worker claude uh code for most of these things? Now the cool thing about claude co-work

[1:11:05] is I mean I use cloud right I've done a lot of automations from a personal productivity standpoint look at all my emails and tell me which emails I should answer I mean and then it does a scheduleuler of hey look at all my

[1:11:17] LinkedIn uh sorry give me LinkedIn post ideas and I've created a skill on how I write and I should be able to push that out now it's great for personal productivity what happens when let's say Suash and me

[1:11:29] are in the same team we have four other people let's say I have um yeah I have Aayush I have Vither I have Ashish we all want to communicate

[1:11:41] you might need a one place where there's collaboration that's happen that's when this agentic workflow thing gets interesting now with claude also I can't easily just share one thing you also see with claude is if I open multiple

[1:11:54] sessions sometimes it forgets what's happened on the on a different session you wanted to have cross cross functional or cross session memory Right. So that's something which agents when you're able to build it's able to

[1:12:06] do a good job. You want you want it to be private. You want to make sure this data is deployed onto your server or specifically let's say onrem. So great for regulated regulated industries right. Um you might want to be like hey

[1:12:20] Google ecosystem. So there's a lot of these nuances that start kicking in from an enterprise standpoint. Uh so that's the deployment side which is hard right? So what you can do however today is build out your

[1:12:33] prototype on lovable and then you can work with AI engineers, AI forward deployed engineers, maybe someone like Elizo to help you take it live. Okay, so I'm going to leave it leave it at that. Uh the TLDDR version today for

[1:12:48] you is your single agents, you have multi- aents, you have an LLM. An agent basically connects to an LM connects to tools, connects to modules, connects to launch base. I showed you the entire infra stack right which is what here is

[1:13:04] the entire stack here is where Liza comes what you will learn you already done the hard part today which is understanding the nuances of studio right and nuances of what encompasses an agent and then after that I showed you

[1:13:19] lots of examples and you'll build something like this today cool can if there are some questions you can take that over. Let's see if there's

[1:13:40] same. Uh I think that the differentiation they're trying to show real estate agent right? So like I think that terminology has got popularized. So which makes use which makes use of LMS following front end and backend

[1:13:54] hallucination in the back end for you actually talk about some of that on his uh without API keys, can you connect multiple agents? You can if you build

[1:14:08] everything within Liza agent studio, you don't need multiple API keys. Uh what if one agent cannot complete its task before passing it to other agent multi? How to fix this issue? You get logs for everything. You get to know where an

[1:14:21] agent has specifically failed. So you'll be able to see okay it's failing here fix our issue and then move forward. So as much as we're low code we have a low code plus a pro code or a dev mode as well to our platform.

[1:14:35] That's pretty much it. S on to you brother. >> Oh well yeah thank you so much any for going through studio and every concepts of agenti in depth. It was really really interesting to uh you know know and

[1:14:50] learn about all those things. chatting for everyone. Uh now time to begin the fun part everyone. So uh if you guys are you know uh opening your phone and just seeing the seminar or lying down eating something it will be the best time to

[1:15:03] just you know open your laptop and uh let's begin the uh actual building part and uh let's take actually 2 3 minutes to you know set up your environment and everything because at the end of one one and a half or whatever time you have

[1:15:17] left uh my goal would be that everyone present here in this uh seminar should be able to say that okay I have built a pretty much complex multi- aent orchestration uh by myself without even uh writing a single line of code. So

[1:15:31] uh writing a single line of code. So yeah uh let me also share my screen. So we will build everything uh along. It's not like uh I'll just be explaining. I want everyone to get involved. I want everyone to uh keep on commenting

[1:15:43] whatever ideas they're going to bring to life. So if you can see my screen uh first of all um let's just open architect. I you must be having this

[1:15:55] document with you. If you are not having it's also fine. I'm just sharing the URL in the link in the chat. Once you open this document, so Anie has right? How you can manually build EI agents, how you can write the system

[1:16:09] instructions, prompt, role, goal, choose the LLM model. But in today's time, I manually, right? So people just want to prompt their way and build something.

[1:16:22] cloud and chat GP and everything, right? So for that specific purpose what we have done is we have automated the entire process of agent development just by giving prompts. So if you have this document uh if you can see this lizer

[1:16:37] architect link I just want you to click here lizer architect. If you don't have the document it's fine you can just click on architect new and I'll also click on architect new and I'll also share this URL in the chat. Here it is.

[1:16:50] So first of all yeah and just keep on informing me that okay we have uh done this we have done that if you're facing any issues I'll be seeing the chat along facing any issues so let's just uh take a moment you know and set it up and then

[1:17:04] we'll start building together once you are in architect.net you will see a you to continue with Google. That is the easiest way. You can also continue with Google. Enter email ID and password. But it's very easy to set up using Google.

[1:17:19] Just enter your email ID and sign in. It's a basic process. I think nothing special here. So, I just want you all to set up it very quickly.

[1:17:37] enabled for you. just fill it up. Let's take 5 minutes to do this. architect. So I won't be needing to do that. But for everyone once you guys are

[1:17:50] logged in, just let me know in the chat. We'll proceed. Okay, I can see done. Continue with Google. Yeah, that works. Cool. I think as soon as you log in, you will see this screen. This is the homepage of architect. First of all,

[1:18:04] quickly to explain like what architect is, why we have built it till everyone is logged in. So you saw the manual way of building agents, right? In the studio studio platform, but let's just focus on architect first. Uh and then we'll move

[1:18:19] architect first. Uh and then we'll move to studio. So studio is where the agents is actually residing, agent is actually living. uh as mentioned by any agent is nothing but where how to make agents is nothing but where you define that okay

[1:18:31] this agent has to perform this role it has a goal which is this it has to have some instructions behind the scenes you can chat with the agent in a studio you can connect tools to the agent in studio you can collect connect knowledge base

[1:18:44] and everything right all this process uh what we have done is instead of doing it manually architect is a platform where you can create end to end agentic applications what do I mean by agentic applications Right? So an application an

[1:18:58] app like a web app whatever you are seeing right now or whatever you are seeing in uh let's say Facebook or Instagram or something that is a normal app right but what is an agentic app so let's say you are building an app that

[1:19:10] has to do a lot of reasoning task that has to do some thinking task that has to take multiple actions so behind the scenes if in any app there are agents what should be the next step there are agent taking actions connecting to Gmail

[1:19:23] 9:00 a.m. Okay, this is what you have to do today. Sending you a summary of uh your work uh like your slack, your Google meet, your teams and everything, right? So that is a kind of aentic application which we can build using

[1:19:37] architect and if you see in the left side there's a lot of things, right? So it's nothing don't get overwhelmed by it. I just want you to click on the prompt library if you can see. So just to show you what kind of use cases you

[1:19:51] just can go through all of the categories that we have here and it is categorized by different fields like marketing people can think of building a marketing team of agents that can write blog post that can create social media

[1:20:05] content, perform SEO analysis and generate graphics for my brand brand. Right? So all the things of all the use cases that you can think of which certain kind of thinking, certain kind of action that has to be taken

[1:20:18] autonomously, architect will help you build those things. You can see a marketplace section which will show you all the cases that people have already built on architect. So we are going to build it ourselves today. This is

[1:20:30] something which you can uh try and check out later on also. Now let's come back to the homepage. Okay. So I just want to check on very quickly. Everything is logged. Everyone is logged in. Okay great. So let's proceed. Uh okay. So if

[1:20:44] you can see this document uh if you have it you can open it in your new tab or otherwise I'll just keep on presenting what we are trying to build. Let me first explain that okay and then we'll start building part. So if you see the

[1:20:56] use case uh there are a lot of ways to build out agents and there are a lot of right you can have a single agent working behind the scenes. You can have working behind the scenes. You can have a manager sub agent kind of

[1:21:10] agent with multiple skills. So there are a lot of ways right. What today we are going to build is a very uh I would say intermediate complex level of agent orchestration which is a manager sub agent pattern. What is a manager survent

[1:21:24] right? It's nothing. It's actually what it sounds. So in a in a company or in an enterprise or in a professional environment you will be having a manager people right and that four to five people will have to perform a single

[1:21:37] specific task which they are responsible for. So that is the same pattern which we are also following in the agent system. Now what yeah right dude when you're sharing you can just close that.

[1:21:49] >> Okay. Okay sure. >> Cool. Thanks. >> Great. Yeah. So I think now it's visible completely. So as I was mentioning what uh is what is the problem we are going to solve right? So I just want everyone

[1:22:02] to you know go through the problem. I am going uh it along with everyone. So first of all what is a SDR right? SDR is nothing but sales development representative and it is what it actually sounds. So you know let's say

[1:22:14] if I am working as a SDR in a company I have to do multiple repetitive task myself. I have to search for prospects manually that okay these are the companies which are relevant to sell my products. These are the companies that I

[1:22:27] customer profile would look like. I will give a ICP score to every uh enterprise that I'm reaching out to. Then I will also have to do a deeper d uh do a deep research on deep research on the people that whom I should reach out right that

[1:22:42] department is very relevant for me and I should reach out to and then once I have decided upon the people I also have content for them that okay this is my product this is what I'm going to sell

[1:22:56] architect right so I need to have context about laser architect first that help you with this that it can make a genic application sent to it and this is how you can use it in your enterprise and then once the outreach content is

[1:23:10] generated I also have to send them the mail the LinkedIn post the LinkedIn emails or uh via slack or via telegram or via chat so any kind of tool that you want to sell emails and this use case is not only limited to let's say if you are

[1:23:26] you want to reach out to multiple recruiters the same process could be followed this the problem statement will be different. So today we'll be focusing more on an EISR outreach assistant and that does the entire manual work of an

[1:23:42] SDR that the SDR is currently doing and uh what the steps would be which the agent would be following is that okay first of all I want to do a research I want to do a research on what companies I should reach out to what people I

[1:23:55] should reach out to the company research and why is this the best time or why is and how should I reach out to those companies. is whom should I reach out to is the next step that okay what are the prospects who I should send the content

[1:24:09] what content should I send it to what the outreach content would look like how I can even draft a mail and then what I can do is I can add a human in the loop that okay once the AI agent has drafted the content for me that this is what I'm

[1:24:22] going to send to these guys let me just uh have a approve button and let me just give a functionality so that I can edit the content myself and then I will reach out uh and then I will finally send the content to the relevant people. Right?

[1:24:35] So all this flow which I am doing manually as a SDR as of now in an enterprise can entirely be automated using agents and just to give you a context of what are the agents will be involved in this process. So we are not

[1:24:49] making a single agent because uh it's so simple and it sounds boring to uh say agent only. What we are going to do is we are going to build entire chain of agents. two three agents will be working systematically. There will be a one

[1:25:04] manager agent which will be handling three agents along together. So if you can see it clearly, let me zoom in a little bit. We have a research agent, right? The entire task of research agent is to search for the relevant companies

[1:25:17] that I can reach out to. Prospect agent will search for relevant prospects inside that company that the research agent has already searched. So the input will go come from the target company that I have entered. Let's say uh

[1:25:29] Accenture is something which I can which I think that okay I can sell my product to. I give Accenture as an input and the research agent will find out the reach out to this company or even this research agent can find out multiple

[1:25:44] Accenture. Inside Accenture whom should I research? Whom should I reach out to? The prospect agent will will tell me. Now once I have the relevant parties, relevant people that I need to reach out. I have a manager agent. This is a

[1:25:57] outreach coordinator. It will coordinate between three agents. Its entire task is to just delegate the task between people like a manager right in the company. So uh this guy is the brain behind the uh flow and these guy will think that okay

[1:26:11] for this particular task for a Gmail drafting functionality I should reach out to Gmail draft agent uh for a personalization let's say I want to reach out I want to reach out and write personalized content. So the

[1:26:23] personalization agent will do this for me and actual writing the uh different writer agent will do it for me. So these guys is manage all these three agent and is to decide which agent should do what task at what time. So this is what we

[1:26:37] task at what time. So this is what we are going to build and uh not we will not actually stop here after building agents what we also do is we'll connect the knowledge base to it. So even if I have to reach out to someone, I have to

[1:26:50] first know that what product I'm selling and what is my value proposition, how my Help the other party which I'm selling to. So all that information we can this information is documented about my product, about my uh company, what is my

[1:27:05] value proposition and etc. So that will provide to the agent as a knowledge base. Obviously we are going to make uh four to five agents in the flow. uh also the agents will have a capability to do a live search in the web right so I have

[1:27:20] existing it can do web search it can do web fetch it can find relevant companies for me and then we'll be attaching a Gmail tool so everyone will be also able to get the drafted content into their emails uh automatically and yeah

[1:27:35] obviously everything is just a prompt to a working UI no coding required nothing a working UI no coding required nothing uh is required there so let's just uh pause for a second here. Um, and if you guys are great, I'm excited. It looks

[1:27:48] good. This is exciting page. Okay. Okay. Great man. I think looks good till Okay. Great man. I think looks good till now. So, let's actually start building. theoretical part. Uh, we'll cover while building. Cool. So, I want you to click

[1:28:03] on architect.ne. Uh, this is the homepage of architect. and how we should tell architect what I have to build right. So that prompt engineering which Ani was talking about that it is a very uh good skill to have how can we explain

[1:28:19] agent behind the scenes which is thinking how many agent I should build instructions will look like and everything right but for you guys and for general users I don't want you to uh learn prompt engineering first and then

[1:28:34] come to architect so architect has the capability to handle whatever kind of prompt that you're writing even if you're writing build me uh AISD agent,

[1:28:46] it will give you a very good result in the first word itself. But if we are able to give it more context, give it more uh context about how the UI would look like, how the flow would look like, what are the functionalities that uh I

[1:29:01] more context you give, the better it will perform like every AI, right? So one way to do is write out a simple line write out a very vague prompt that okay it. It will ask certain questions for you. It will build out the entire

[1:29:15] applications for you and then you can keep on iterating keep on refining the final app accordingly. Other ways to in the first initial go just tell everything I would say just have a chat with uh claude or chat with chat GPT or

[1:29:29] even you can have a chat with architect to help you organize your thoughts that okay what are we actually going to build if you have a clear document written down that this is what I'm going to build which I have myself written it

[1:29:42] down so I'll just go through it once it is very simple the same thing which I outreach agent now how to write a good prompt I'll give you like few tips uh while working with architect but uh

[1:29:54] doesn't matter like if you can even uh write a very vague prompt it works the same way you just have to iterate a little bit more on top of it but if you can break down your thinking in in these sections first you can define the goal

[1:30:06] what are you trying to do so let us spend four to five minutes you know uh just reading the prompt because I think this is a good skill to have not just anything which you are even working with cloud even working with chat GP or some

[1:30:20] other platform right so break down your thought process uh if you are trying to five things I would say if you're building an agentic application specifically give the agent the goal that you want to be achieved right first

[1:30:33] of all tell us in very briefly two to three lines okay given a target company the user want to sell into research that company identify the best prospects to reach out to and draft personalized outreach all grounded in the user's own

[1:30:46] company details from an uploaded knowledge base the app runs as outreach. By the way, I have not written it manually in this uh complex manner. I just had a chat with cloud or I just had a chat with Elizer agent actually and

[1:31:00] ask it to give me a detailed uh prompt which an agent can understand. So if the which an agent can understand. So if the language is complex it's not my uh writing the agents is the next part which we have to define. If you are very

[1:31:14] agents you can include this section. Otherwise I would say if you don't know or if you don't have an idea about how many agents I want, how many uh what are actually skip this part. It's just still fine with architect. But since here I

[1:31:29] agent to research. I want a prospect agent to do a company research and inside that company find out prospects for me. Write personalized agent. Uh one agent will be actually writing the content. there will be a writer agent.

[1:31:43] So I myself know um that okay I need this four to five functionalities. If you are not very comfortable with defining agents I would say define the functionalities that you want to achieve. So research agent forgot about

[1:31:55] this text right you decide describe the functionality research as the target industry size whatever parameters that you want to research on u using the you define the functionalities what you want to do and then what inputs is going

[1:32:11] have to provide a knowledge base you have to provide a company context of what product you are trying to sell uh let's say uh some months uh later you

[1:32:23] are starting your own startup you have a uh drink product to sell right so you have to provide the specific details of what kind of drink you are selling uh helpful how it is healthy and why should you should consume that drink. Right? So

[1:32:38] that everything we have prepared in a document that okay these are the uh details about my product that I am selling. Inputs may inputs inputs I will I have in mind that okay I need to sell to this company I need to sell to that

[1:32:53] company I can specifically mention the name of the company otherwise I can ask agent to uh fetch the company do the company research also for me. I can also define the flow that okay how the flow would look like what are the steps that

[1:33:06] First of all, research about the company. Find prospects, generate outreach, send the email to them. Right? It's pretty simple. Uh what output I want? I want a company research summary. I want suggested prospects. I want a

[1:33:20] drafted email. Uh and maybe I can also get two three hook lines, you know, to make the end user open my email. Then the last section would be how you want simple. I just wrote down that okay, give me a single screen, two column

[1:33:35] should look clean and clean clean and professional. Uh these are the three to four uh sections that will be in the UI and you can describe very uh briefly or very uh in a simple language that okay this is how I want the app to look like.

[1:33:50] Now once you have this prompt ready with you or even if you don't have this prompt ready with you it's fine. Uh but since we have this uh just copy this since we have this uh just copy this prompt uh if you have this URL just copy

[1:34:03] is we going to paste the same prompt in architect and once you have paste the same prompt in architect just hit enter and till it is loading if you guys are have any questions have any doubts feel free to ask me.

[1:34:18] >> can I specify? Yeah >> everybody has that link so just place that again here no your uh GitHub link. Okay. I think yeah, I pasted it while Okay. I think yeah, I pasted it while starting. I'll paste it again.

[1:34:33] >> Yeah. Yeah. I think Ana shared it before with everyone. But yeah, I'm pasting it with everyone. But yeah, I'm pasting it again. Also, Suy, while you're showing this, I think from this is kind of I feel like going to step three if you can

[1:34:46] also show the 0ero to one just architect like put a small prompt. How does a plan >> yeah, that's what we are going to do now. Yeah, correct. Cool. Yeah. Yeah.

[1:35:02] the prompt for everyone, what architect will give you is first of all a set of understand more about the use case that you're trying to build. It want to know that okay these are this is the it it already knows that this is the app or

[1:35:18] this is the UI that this is the agent that you want to build. But I need few more context about uh what you are trying to build okay to help you make the app in a much better way. So you will see a set of questions here that

[1:35:31] okay first question the knowledge base and by the way this question set will be different for different people uh because behind the scenes an agent is thinking that okay what question I should ask so it can vary uh depending

[1:35:43] given. So whatever questions you are getting in your screen just read it out and just select the relevant option that you feel is correct. For me it's like the knowledge base how should the user's company do be handled. So let's say I

[1:35:56] users upload the doc each time they start a new outreach flow. Second question is what should the research agent lean on for live company info. So it is asking me for live search what I should want what I want to do. So

[1:36:09] perplexity models are quite good in this kind of functionality to do a realtime perplexity pro. And third question is will this app need to save and manage data across sessions. Uh if you need a database if you need a authentication

[1:36:22] and if you need to persist the data you can select yes otherwise for this use authentication. I'm only going to use this app for me. So I am selecting no. And fourth option is use default laser agents to build this or build this using

[1:36:36] g agent. So g agent is a completely separate way of defining agents that we will be focusing uh later on maybe but for now I would like you to select use default laser agents and lizer agents is basically the same agent that was inside

[1:36:51] showed you guys right so these are the set of questions uh which architect will ask you everyone will get different questions so don't worry just read the question and select the best option possible uh for you and best option that

[1:37:04] looks to you and once you have done this hit continue hit continue and till it is uh building till it is generating the next steps. I would like you to tell about the flow of how

[1:37:18] architect does the entire work. Right? Once you give it a prompt, once you define a prompt, once you decided on use case, you give it a prompt. It will ask you few questions that okay, how uh you want to envision the app. Uh these are

[1:37:32] want to understand more about. After that, it will actually create a product requirement documentation for me which I can see in the screen as of now. So after I gave the prompt, this is the plan. This is the PR architect generated

[1:37:46] itself and we are going to go a little bit deep in this. After the PRD is generated, it will proceed to generate the agents for me automatically. So this is the plan section which we are currently in. Then we'll move to the

[1:37:59] agent section and then we are going to go to the app section. Uh so the flow is once I get the prompts once I get the questions answered I'll get a detailed build out the agents defined in the product requirement documentation and on

[1:38:14] top of the agents I will build the UI so we'll we're going to go through the entire flow but uh if you guys are okay till here let me know if you have pasted questions architect has already asked and if you can see a plan screen like

[1:38:28] this how does it know when is the rack proc process the internal document to be used. Okay. So whenever if you remember input right in the input I have specifically mentioned that I will be

[1:38:40] sharing you the details of the company that I want to propose right so if you that okay I will be sharing some information I'll be adding a document automatically create a knowledge base for me and then we can add the document

[1:38:55] there. So yeah, two things. >> Yeah, >> people are issue with authentication. >> And second is I think someone's saying we shared the uh the credit with them.

[1:39:09] No, someone's saying it's paid. It's asking premium. >> Uh no. So yeah, let me also clear that out. So first of all, uh the issue with login. So there are two ways to login, right? One is login with Google and the

[1:39:24] second way is to login with email ID and password. So let's say if you're facing can try the different way the another way with login with email id and way with login with email id and password that will work for you. And if

[1:39:36] can just sign up with Google. If that is not working or you can enter your email, enter your uh organization name, you can keep it anything and uh password and confirm password and then just read the caption create account. This is another

[1:39:49] way which you can try out to create the account for you. And as soon as you log in, you will actually get $20 worth of free credits. So you don't have to purchase anything. And once those uh $20 is exhausted, we can actually top it up.

[1:40:05] We have a coupon code with us uh in the document which have been shared by you. So I'll be giving you more information about that. But I think for the initial fees for the first prompt that will be done in the $20 that has been provided

[1:40:17] to you. So if you see if you click on uh your login if you see the homepage of the architect here if you are able to see dollar20 that is the amount of currently. So as soon as you log in you'll get this. Yeah.

[1:40:31] there are people. >> Sure. Sure. >> Yeah. So I think that's fair. What you're saying is absolutely right. >> Uh yeah just look at the chat. >> Yeah. Uh if you are seeing zero, if you

[1:40:43] are seeing something, click on this refresh button because uh we have also joining the seminar. So if you can click on this credits, if you see this refresh button, just click on that refresh button and let us know if you are able

[1:40:58] to see the credits there. to login with so that they get the credits or it comes to the it come it

[1:41:13] uh who should be logging into architect will be getting uh $20 worth of free credits. >> A lot of people are show saying it's >> Okay, we can do one more thing then. uh for the people who are u seeing zero

[1:41:28] for the people who are u seeing zero grids if you have this document with you the link which I shared I'm going to share it again so what you guys can do is you can actually scroll to the end section there

[1:41:41] is a coupon right if you see the last section of the document and I'll also just share the coupon code in the chat so that you guys can

[1:41:56] >> We have JJ who I think got the $20. JJ, if you want to share also what you did specifically, my >> Yeah. Or you can refresh the entire tab once. Uh if that specific refresh button is not working, but I think most of you

[1:42:09] should be having the credits with you. I do have Okay. do have Okay. I have the Okay.

[1:42:21] >> Refresh. Okay. Okay. >> Just look at the previous one. It says >> Just look at the previous one. It says detail organization service unavailable. >> Uh but I get the error. Is this error coming on clicking the continue with

[1:42:34] Google? Uh subroy or have you signed up with email ID password? If you can tell with email ID password? If you can tell me that.

[1:42:46] apply coupon code as of now because that will override the $20. So if you are able to see $20, let's first use that. Let's first exhaust the $20 and then we'll apply the coupon code. So let's

[1:42:58] we'll apply the coupon code. So let's keep it with you.

[1:43:11] apps. Okay, I would say just do a hard refresh command shift R on Mac or control shift R on Windows for the entire uh tab of Chrome or Firefox or whatever browser that you are

[1:43:24] just you know they're saying where to put the coupon code. Um what sorry where to put the coupon >> Yeah just see that. >> Yeah I can actually share the way of u

[1:43:38] using the coupon code also. But for the guys who are using who are seeing guys who are using who are seeing already $20 added uh so no need to add the coupon code uh for those guys. But whoever wants to use the coupon code uh

[1:43:51] at the starting once you if you can see my screen everyone once you click on this credit section and click refresh if you're seeing 20 uh no need to use a coupon code but whoever are seeing let's say zero credits you can do a top up

[1:44:06] here if you see this top up button once you click on topup and there is a plan and topup section on top right right you can actually click top right right you can actually click on the right side topup section and once

[1:44:20] you click on this topup section in the left side you will see $150 credits you left side you will see $150 credits you can actually click on buy pack here once you click on buy pack you will be redirect to a payments page and here you

[1:44:35] will see the add promotion code info if you click on this add promotion code you click on this add promotion code here is where you apply architected 50

[1:44:49] applied it so it will not work for me but it will work for you guys and also this information is shared in a video in this URL which Ana might have shared with you guys or I'll just share it once more. So the same video you can follow

[1:45:04] to apply the coupon code. Let me share it once more. Again the flow is simple. You go to architect homepage. You click on the top

[1:45:19] right credit section. You click on refresh. If you see $20 already added, no need to add coupon code as of now. Save it for later. Otherwise, just click on top up. If you click on top up, you will see a screen like this. Or you will

[1:45:32] see the plans screen. Plans screen. But I want you to go to top up top right. Click on buy pack for the $50 credit section which you can see. Once you click on buy pack, you will just uh get the payments page where you can click on

[1:45:47] add promotion code and just enter architect 50 might be uh getting some information to enter your email ID, email id and uh

[1:46:00] address that you can do and then you will not be needed to uh add any credit card or uh make any payment. You can just click on continue.

[1:46:14] see you can follow the same uh I'll just see you can follow the same uh I'll just do a full screen of this

[1:46:26] you like your name and address you can just enter your name and address click on complete order no need to add any credit card or uh payment information and then yeah uh you will see the credits added to

[1:46:41] your account, the $50. Let me read the chat. Worked now after doing control refresh. Thank you. Okay, screen is not visible. Uh I bought the pack still zero. Just uh do a hard refresh guys, I would say. I mean

[1:46:56] a hard refresh guys, I would say. I mean that would solve most of the problems. coupon is cancelled. Okay, I bought the pack still zero.

[1:47:18] Okay. Uh for the guys who are still facing issues or still not able to uh get the get the coupon code and get the credits. So let me know like who who are how many of you are not able to get the credits.

[1:47:34] credits. [snorts] able to still get the credits,

[1:47:48] you can try. Uh but I would say log out login first and do a hard refresh. It would should reflect already. But if even that is not working uh yeah just watch my screen. I'm sharing the other way around if you guys

[1:48:01] sharing the other way around if you guys want to add the credits. So if you guys want to add the credits. So if you guys can go to studio.lizer.ai. can go to studio.lizer.ai. Let me share the URL for that also.

[1:48:20] or in the document also we have the URL given LISER agent studio. You can click given LISER agent studio. You can click there. This URL will redirect you to the riser agent studio and you can actually

[1:48:35] sign in in studio using the same account which you have used to sign in with with Google. If you have already logged in to architect with the same account here also you can login. Otherwise if in architect you have used your email ID

[1:48:49] password you can use the same email ID password and login with email. Once you have logged in in studio by clicking login with Google or clicking login with email, uh you will see a homepage like this.

[1:49:29] My screen is not visible is it? Yes, the screen was visible. Okay. So, if you guys can see a studio screen like this,

[1:49:41] you will actually see a profile. Your profile here in the bottom left section. Your name will be written which you have used to log into studio.

[1:49:55] actually click on this slicer. So there are multiple organizations. So if you password so you will see organization like this. Let's say swish organization. And once you are in your organization so I

[1:50:10] am in basically lizer organization. You should be able to see the credits here. Uh so for me it's showing 89. You should be seeing $20 or $50 depending upon the be seeing $20 or $50 depending upon the credits that you have used.

[1:50:29] the credits here uh after logging into studio.

[1:50:45] here, what I want you to do is click on top up now. In studio, you can click on this small top up now button.

[1:50:58] button from here also you will see the $50 credit addition option. So homepage, click on this credit section. Click on top up now. Uh $150 buy now. And I am in a different org. Once you

[1:51:13] click on this buy now button, you will be redirected to the same payments page like this where you can apply the coupon code again

[1:51:33] everyone or few people are uh able to log in get the code and get the credits change the org to your particular account org. So if you have logged in

[1:51:47] with Suesh, you should be in Suesh organization.

[1:52:03] Yeah. So Lizer or you have to switch to uh your particular organization.

[1:52:20] the issue. Yes. So you should be using the same email guys. So uh don't use the same email guys. So uh don't use multiple emails. So

[1:52:37] Let us just uh work out with the credits. Okay. also so for uh people who are not able to uh you know get the credits and uh you can actually uh go through what I am presenting and you can actually check

[1:52:53] what I'm trying to do what I'm trying to build uh that is also one way to follow along because I'll be sharing the entire stepby-step process myself but few uh cases which you should be remind keep in mind uh for the credits is uh just use a

[1:53:09] single login single email id password the one which you have logged I'm using studio the keep the same ID password in architect also architect also [snorts]

[1:53:23] organization you have to select your own organization once you have logged in

[1:53:43] Once you are logged in, you have to click on this button and then you will are a part of. So let's say if you are seeing Lizer or some other or either choose my organization or choose your names organization that will be the

[1:53:58] names organization that will be the account uh for you.

[1:54:18] Command shift R on um Mac and control shift R on Windows.

[1:54:30] first app. Uh no actually whenever you logged in uh whenever you log into architecture whenever you log into a studio the organization will be created studio the organization will be created automatically.

[1:54:51] Yeah. So okay for people who are able to get the credits I would say uh follow along me. for people uh who are not able to or facing some issues. What we can do to or facing some issues. What we can do is uh uh we'll keep a track of uh the

[1:55:04] users list. So whoever have not redeemed uh the credits maybe we will just provide you the coupon code so that you can uh try at your own time also that will also uh work. Okay. What shall I aail after I use the

[1:55:20] free credits? So that's what the coupon is for. uh $20 uh you will get free when as soon as you log in. And once you have logged in and exhausted exhausted those $20 credits, you'll be using the coupon code which will give you additional $50

[1:55:35] code which will give you additional $50 of credits. entire flow going through uh in the screen which I'm sharing. So you just

[1:55:50] follow along me, you just try to work with me whatever we are trying to build with me whatever we are trying to build and you will get equally understand and uh the credits part I'll just share the coupon code and the for

[1:56:06] what whoever guys are not able to use that coupon code uh you can I'll just [snorts] then you can maybe log in afterwards and then try it out.

[1:56:21] in who have the credits with you and who are able to give the prompt to the architect and who are able to answer the questions who are able to generate the PRD uh just give a thumbs up in the chat so

[1:56:34] that we know that people have reached till the PRD people have reached till the PRD section. Uh

[1:56:48] Please refresh it. Uh what is the next step after pasting the what is the next step after pasting the prompt? Okay. Okay. Yeah, let's proceed ahead also. And for people who are facing any issues, follow along me in

[1:57:02] the video in the screen sharing which I am doing because I think you will get to know the step. from here and then maybe you can try later on also because the steps are always the same. So just understand the

[1:57:16] flow first. Yep. And PRD is as I mentioned a product requirement documentation uh which architect generates itself. And if you can see here it has certain sections in it right. First is the overview user

[1:57:32] stories agent architecture is the most important one. Then we have files website provided by user user flow integrations required UIUX specifications. So what is the difference between um the initial prompt

[1:57:47] product requirement documentation right. So first of all it is a very in-depth documentation of what the actual product will look like what the actual functionalities of the app would look like. what agents we are going to use

[1:58:02] this in this particular app and how the user will use this app. So if you read this PD in a little bit detail let's just go through the agent architecture section here if you see there will be a agents table. This agents table is

[1:58:17] agents that architect will be creating for you. If you see here there are certain columns written agent type agent name description tools trigger provider model temperature all the parameters which I was talking about in lizer

[1:58:33] studio right so the difference that architect has created is you don't have to manually decide what agent you need to create what description you need to write in the agent what tool you need to connect to the agent what would trigger

[1:58:46] which model you need to uh add to the agent what specific temperature and top you need to put in the agent. Architect itself has decided that okay I would have one independent research agent that would research the target company via

[1:58:59] live web search overview industry size recent news funding hiring signals and inffort business plan. So I just want you to take uh five minutes to you know go through the PRD and understand what architect is trying to create.

[1:59:14] So the second agent if you see here is the prospect agent. By the way, this PRD might be different for all the users, right? Might be different for different users. So, it is never a very deterministic flow because behind the

[1:59:27] scenes architect is also an agent which is having an LLM working behind the scenes and that LLM will always generate nondeterministic output. But at the end, the functionality of the app will always be same. Someone might even see four

[1:59:42] see two agents instead of three. So it is completely dependent upon LM of how they want to achieve the task. To right you can follow a path A to B or you can go to A to C and then to B. So

[1:59:56] the end result will be same the path might be different for different people because the LLM is nondeterministic in a way and it will give you different outcomes different results every time. Even if you paste the simple or a single

[2:00:10] time. After you see the PRD, if you see the agents table, you can also scroll down, you will see the user flow. In user flow, if you read it, you will have how the user will interact with the app.

[2:00:25] Just to give you a few points, user lands on the single screen work space. placeholder. So it is telling you that okay once the user launch the app what will the user do and how the user flow would look like. The left column will be

[2:00:40] your company card. The left column will be your target company card. User types company. In the right column you will see a loading state. The prospect agent auto runs. Step two step reveals three for selectable persona. A primary

[2:00:54] So all the things that you will be having in your app that you will be uh doing in your app if at all there are any integrations required with the app how the UIUX would look like. So this is a very in-depth document of the PRD

[2:01:08] documentation of the app that we're trying to build. And if you see towards the right of plan there is also a app mockup. What is this app mockup? Now this is not a not the final app that you're trying to build but a mockup UI

[2:01:23] of what the app would look like that architect will finally build itself. So this is how the final app would look like. We are kind of giving a mockup to you before even building the app that okay your app would look like this. If

[2:01:36] okay your app would look like this. If you like this UI, if you like this uh section what we are presenting to you, you can continue or you can actually change before only that okay this is not what I would like. I would need a dark

[2:01:49] what I would like. I would need a dark mode in my UI. Uh so what architect can in architect can be done just by prompting it. So in the left side here is where the chat would reside and all the interaction that you will have with

[2:02:03] architect will be we will be doing in this left chat left side chat bar right and if you see in the bottom left you will get a pop-up like a router agent plan is ready. If you want to start build you can start building. If you

[2:02:18] click on back to plan mode and you can actually ask architect to do any changes in the plan. If you let's say want to add one more agent, if you want to add one more section in the UI or if you want to remove anything from the UI, you

[2:02:31] can just come to architect and prompt it. So if you guys want to try it, you can also try let's say I want the UI in a dark theme.

[2:02:43] uh for people who like the light version uh for them you don't even need to run this prompt but uh for someone like me I need uh my apps in dark mode so I just come to architect and I tell it that okay I didn't like the light version

[2:02:58] that you have given me I want the same thing in the dark version so every prompt or every iteration or every change that you want to do everything could be done just by giving architect a prompt and architect will be able to do

[2:03:12] the modification for you and in real time you will see the changes and you will actually see how the architect is working behind the scenes and yeah it has changed the UI for me if you can click on the app mockup you will see a

[2:03:26] dark version not the light version so and this is just a mockup by the way this is not a functional UI we'll be getting the functional UI at the end so yeah and now it looks pretty good to me and

[2:03:39] now it I'm satisfied with the UI architect is giving me. So I can architect is giving me. So I can actually click on start building here. Start building what it will do is uh and if you guys are okay till here just

[2:03:51] click on start building. Uh what start building will do is it will tell you that it is proceeding to build but to build out an app you will need certain trying to build and how the app would look like. So we are passing certain

[2:04:06] trying to build. We are passing what? that target generated. Then we are passing certain artifacts. Artifacts does artifacts is not anything but the app mockup which we just see that okay

[2:04:20] just passing this as a context to the final app to the agent that will be Along with this I'm passing certain skills that it should be able to do uh take a knowledge base and this artifacts might might differ for uh

[2:04:36] other users. Right? So whatever plan artifacts you are seeing you can select everything by default everything will be selected and you can click on build with this. So once again just to go through the process. We give the prompt. We

[2:04:49] answer the questions. We did a little bit of iterations in the mock UI. We changed from light mode to dark mode just by chatting. Uh when we click on back to plan mode, we chatted and uh the UI got changed. After the UI got

[2:05:03] changed, I just click on start building. I see this uh popup where I'm selecting what context I want to give to architect and uh I do not see those two skills. and uh I do not see those two skills. Yes. So it is not necessarily to say

[2:05:16] that always these skills will be required. As I mentioned everything might be seeing a different UI. I might be seeing a different UI. Some other person might be seeing a different UI. Someone might be seeing a different UI

[2:05:29] or dark mode by default itself. It's an LLM working behind the scenes and every time you ask something the answer might vary um even if you give the same prompt. So that is actually the beauty of it right this is this able to think

[2:05:43] in real time and build out something in real time it will not always be following the same process the deterministic process so let's proceed to the next step once you are here I would like you to click on start

[2:05:55] would like you to click on start building and build with this now the actual uh building of agents will start happening and it will start building from your plan it will load for a few seconds and you will be redirect direct

[2:06:09] to the agents screen from plan screen and once you're in agent screen yeah it will take few uh minutes to process you can let it load completely and yeah for

[2:06:22] people uh whose uh all my chats are lost due to network please share the code again so I have this document where you can copy the entire prompt and paste it

[2:06:35] again you can enter architect in a new tab architect new is the link link and you can paste the same prompt you can begin again. begin again. Yeah, also focus more on uh learning how

[2:06:47] we are trying to build the agentic applications. So it is not very much restricted to this particular use case. Uh why I'm saying this because uh every person is having a different use case to a different problem statement, right?

[2:07:00] You might be having a problem statement that you want to automate uh LinkedIn different problem statement where you want to have a personal assistant to uh fetch all the emails in the morning and tell you what are the pending items that

[2:07:14] pending actions need to be taken by you. So every person has a different problem statement. understand the process that we are trying that that we are following and then the process will always be same the change is always the initial prompt

[2:07:27] that we have given. So try to uh work along with me try to uh be with me and understand the process that we're trying to build. [snorts] And if you can see my screen, if you scroll down the left sidebar, once we click on build agents,

[2:07:42] it started to build agents. But along the way, it is also showing me what it is doing all the live activity. So it says that okay, I will analyze this PRD and build the EISD outage assistant. Let me start by setting up my task tracking,

[2:07:56] defined certain tasks for me that okay these are the three task which I would be performing. It is actually created a knowledge base for me. Uh it is also creating the agents for me. You can spend few minutes reading the chat also

[2:08:10] till the agents are getting built and it will take five to 10 minutes depending upon the uh LM and depending upon the prompt to build out the agents and to build out the app. So meanwhile try to read everything that it is uh doing it

[2:08:25] here. So that you can understand how architect you can understand anything which is going wrong. If you can understand what architect is doing currently, it will be very uh realtime communication happening

[2:08:38] between you and architect while it is building things and while it is doing the things right. So yeah if you can see and if you are So yeah if you can see and if you are able to get some agents on your screen u

[2:08:52] let me know if you are able to reach till this point. If you are in the plan mode, if you are in the agents mode, uh some people might be uh in the app some people might be uh in the app itself. Uh agents mode, great. In build

[2:09:06] scenes. So, prau right now we are using cloud only behind the scenes. But it is not uh just a model. What we have done is it is using opus 4.8. But what we have done is around the llm we have kind of provided or created our own hardness.

[2:09:23] Harness is nothing but how the LM should perform within a given environment. What actions it can take uh what tools it can do, what prompts goes behind the scenes, what skills goes behind the scenes, all those things. Uh we have written it down

[2:09:36] those things. Uh we have written it down among the LLM around the LLM and then the LLM is able to perform the action for me, perform the agent creation for me, write the in instructions for the agent, uh decide what agent to wait, how

[2:09:50] much agents to create. So all those kind of stuff right and u actually it is a lot going behind the scenes. So architect if you want to understand how architect is working uh it is a very complex orchestration of agents only

[2:10:05] which we are doing behind the scenes. Uh and you will be having a lot of and you will be having a lot of fundamental things to make it work and lot of hardness issues. you will be having a lot of uh agent orchestration

[2:10:18] pattern that you need to read about uh to get a product like this working to get a product like this working right. So if you can see uh it will automatically go to the app section once it has built out the agents. You don't

[2:10:32] need to even touch architect uh while it is building the agents and while it is building out the entire app. You'll just have to hit enter in the starting part and then leave the app till it is building the complete section and in

[2:10:46] to understand how architect is working or how agents are working behind the scenes how multi-agent orchestration is working behind the scenes uh so you will need to have a in-depth knowledge of agentic AI and multi-agent

[2:10:59] orchestrations right uh and I think simply none has a proper course for it simply none has a proper course for it also so if you can search for There is a Agenticia in multi- agent system. So

[2:11:12] they actually teach you uh about how to build things like this which will also clear a lot of concepts for you uh that how architect is working, how studio is working, how agents is working because let's say if you are writing any code

[2:11:26] you need to have the base clear first of all that how even AI is working or how even agents are working. So in this proper course which you might uh do a research about yourselves but uh they have certain basic concepts covered in

[2:11:40] this course they have certain demos which they are trying to show it to you guys just like the demo which I'm giving it to you but this will be a bit more can't explain or we can't explain in this two to three hours of session. So

[2:11:54] we have this uh kind of a 10week course which you can try it out and which you you can join the program. They will be having uh 40 plus demos covering seven real world projects, all the skills related to agent, multi- aent system,

[2:12:09] rag, MCP, planning systems which I just talked about and lizer is one way to do things. We have other platforms also. So these course also covers platforms like lang chain, QAI, autogen. So you will have a overall of idea of

[2:12:24] you know how the process how the flow is working behind the scenes how agents are things which you can learn from this course is okay let's say even if I have created the agents how can I deploy it how can I add tools to it uh how the

[2:12:40] agent is working how can I basically uh think about adding tools knowledge base all the basics of LLM what which model which I should I choose and where should I choose that all the multi- aent ecosystem along with realtime projects

[2:12:55] that you are going to build. [snorts] So yeah, you can explore this yourself. I'm just uh telling so for someone who wants to understand more about how things are working behind the scenes. So this is a course which you can try it out. You can

[2:13:07] will be learning and these are the skills which are very much in demand uh in the market also as of now. As you can see all the tool integrations, what are vector databases? See, what is semantic search uh what is context engineering

[2:13:21] because prompt engineering also a thing of a past. Now we are now moved to context engineering. So all the things related to uh AI skills tools uh they are covering and uh obviously you'll be getting a certificate at the end. So it

[2:13:35] applying somewhere that okay I have already done this ex extensive course these concepts where I have built out demos uh where I have made a capstone

[2:13:47] project and uh yeah you can see few of them a few of the reviews which already uh people who have tried out this course have uh already tried and found success you can search about the impact studio so the same way which uh

[2:14:02] architect is working The same way which you which I'm telling you that okay this is how architect is doing the creation of agents creation of UI the same things which also you can learn here that okay but a lot in depth

[2:14:16] a lot more detail a lot more conceptual wise how things are working behind the scenes and uh here I am available you for you to explain you things but we know answer every question so in this

[2:14:30] program you'll also be getting a lot more people from team who will be taking your queries will be helping you prep for any interview, prefer any assistment and uh they will also be providing job opportunities. So just wanted to give

[2:14:42] you a little bit context for the people who are more interested who are more behind the scenes how agentic AI is developing how to get more around agents multi-agent systems and everything

[2:14:57] and till we were talking uh if you can see my screen architect has actually built out the entire application for me. Uh for some of you it might be still be still be building for some of you it might have done building few minutes

[2:15:11] might have done building few minutes ago. So let's just uh have a comment in chat for people at which phase you are currently still building or in plan mode or in build mode or in app mode in agents mode what you can see let me

[2:15:25] agents mode what you can see let me know. [snorts] uh the course free structure also I can

[2:15:37] share you for the people who are more interested. So uh we have actually uh exclusive workshop discount for the people who have joined and uh I think Ana can also uh tell more about the fee structure and everything but for the

[2:15:53] people who have joined this workshop we exclusively have a 35% discount. I mean simply learn exclusively have a 35% discount. So the course the in-depth course which you'll be learning for over 10 weeks around 2 months which is a live

[2:16:07] realtime session where all your doubts will be cleared you'll be making demos we'll be having some real world prototypes uh the original cost is around 1199 but specifically to this workshop we are

[2:16:21] but specifically to this workshop we are giving it at a discounted rate of 78k and uh obviously I think you should be researching a little bit around yourself researching a little bit around yourself that okay is this uh uh is this relevant

[2:16:33] for me and uh how much help I can get from this course because the topics they're covering is very relevant in today's market. So I think uh for people who want to actually learn how things are working because prompting is one

[2:16:46] thing and getting the app built is one thing but that process can be a lot more easier and faster if you know how behind the scenes everything is working how agents are using LMS how tools are getting connected what is rag what is

[2:16:58] getting connected what is rag what is MCP what is uh workflow automation what right [snorts] should I deploy or wait uh we can wait uh deploy is the last part So I think you are ahead of us. Uh Prau

[2:17:13] think you are ahead of us. Uh Prau can show slide number 564 once. Okay. 56 can show slide number 564 once. Okay. 56 is this one. Yeah. Uh yeah, these are all the learning path that they are covering in the program

[2:17:25] that they are covering in the program deep dive.

[2:17:40] How much is a multi-agent system certificate in USD? Um, so Ana, you might be able to help there. Uh, >> yes sure. >> yes sure. >> Yeah.

[2:17:58] I take 6 minutes. Great. My agent is available. Great. accommodate distance in US timing wise? I think we at the end we can also uh you know take all the questions related to the course and have a Q&A for everyone

[2:18:13] who is having any doubts related to the course. We can take up agents mode build mode still building. Okay. So I would say let's just wait till the app is building. Uh I think we can also take a 10-minut break uh now because I

[2:18:30] think we have been in session since more than 2 hours actually. So let's take a 10 minutes break I would say and uh let architect build the entire app uh and

[2:18:42] then maybe we'll pick it up from there. So meanwhile let the questions coming in on the chat. I'll be happy to answer the questions till you are waiting. If you guys want to eat something, drink something this is the time to do.

[2:19:12] after credit just give us steps of plan and deploy. Yeah. So actually uh I am and deploy. Yeah. So actually uh I am again pasting the document link here where we have mentioned the prompt. So

[2:19:25] also don't think of it like a single use case prompt that you are trying to build. I mean I would say I would actually prefer people if you can think of a personal pain point that you guys are facing in your real life. I think

[2:19:38] then you will feel the actual worth of the product and actual uh high you will get I would say if you are if you are able to solve a problem which you are able to solve a problem which you are facing in your uh daily life [snorts]

[2:19:57] Sure, we'll uh just look at all the errors also uh so also little bit to you know give you more context around errors. So if you have tried out platforms like replete lovable or even claude or even open or something

[2:20:14] it it will not always be a 100% success. I mean uh just to be very honest about I mean uh just to be very honest about the LLMs and AI if you are getting something is not working as per your expectation that is bound to happen uh

[2:20:30] even if in this prompt it won't happen in certain other prompt uh that error might come. So you it's our job that okay we first of all have to give the proper context to architect about what it is trying to build and uh not even

[2:20:44] with chat GPD because even if you work with chat GPD because even if you work with cloud code not always the proper uh be having uh when you started working

[2:20:56] with the app right but for what for architect what we have done is let's say if you are facing any error or if you are facing any issues with the app or thing you can actually take a screenshot of that uh error take a screenshot of

[2:21:10] the problem that you are facing uh so let's say while using this proper particular app I see error in the UI or I see some kind of thing which is not working I can actually take a screenshot of this uh app I can actually paste that

[2:21:26] screenshot in architect let me just show it to you let's say while using this app so my app is already built and working. But for some people who might be facing some issues, I paste the screenshot there and I ask

[2:21:41] I paste the screenshot there and I ask architect to that okay fix the error architect to that okay fix the error that is coming up in the screenshot or the app. So I am specifically showing architect

[2:21:54] specifically telling architect that okay can you fix this up? So now it will be able to fix that error up. So by nature LLMs are very nondeterministic. So some thing might break because behind the scenes it has to write a lot of code. It

[2:22:08] has to build out agents. It has to build out the app for you, right? And if it might break somewhere or the entire agent might break somewhere.

[2:22:26] share the coupon code, we'll share the u URL and everything. So people who are not trying it right now, they can certainly try it anytime they want. Uh it's actually public. Uh the architect new is public. You can come to the

[2:22:39] architect. Anytime uh we sharing the coupon code. Um and most of it is actually uh the login issues most of it will go if you can also try it in incognito because a lot of you guys might have uh opened architectural

[2:22:53] email that might also be throwing you the error.

[2:23:27] Okay, we'll begin in uh five six more It's

[2:24:05] know share your experience while building. So uh we went from prompt we building. So uh we went from prompt we went from prompt to agents to app uh

[2:24:17] and uh after the break we are actually going to cover uh the agents how it will look the prompt the rules the set of instructions that architect has created the app that it has built but let us know uh how you find it till

[2:24:32] now I any kind of uh feedback while building out

[2:24:49] you know tell me S like what uh difficulty did you face?

[2:25:19] understand uh that it is it is not claude that we are talking with and it is uh kind of giving you a response. It is not like a chat agent or a chatbot is not like a chat agent or a chatbot that okay I want this answer and uh I am

[2:25:35] prompting this so give me this. uh there is a lot of uh things happening behind the scenes right which we will be covering in the next part that [snorts] architect has to understand the context built out the AI agents. So AI agent

[2:25:51] building itself is a next level of complexity that we are solving here. uh because even if you try to build out AI agent using langen and lang graph you will have to know a lot of technical things technical stuff uh around lang

[2:26:05] around AI agents around LLM models uh you would need to know how to code then you would need to know how to code then only you will be able to make it work um but that entire process of building things manually uh that we have kind of

[2:26:20] you know automated using architect just by giving it a prompt and on top of it on On top of building the agent, we are also providing you the app layer and that app is connected to the agent. So it is a lot of complex things happening

[2:26:33] behind the scenes. In the final result, you will only get to see the simple uh you will only get to see the simple uh or the uh app UI which you think uh is very uh looks easy and looks very easy to build out. But uh behind the scenes

[2:26:47] are working and the agents are getting created automatically. created automatically. [snorts]

[2:27:04] I'm 50 plus. Any carrier for me in this? Um yes. I mean I don't think age matters um in today's world. Even if you are 60 or even if you are in uh 16 year old kid I think if you find it interesting you can definitely figure something out.

[2:27:21] I can use this as a tool to learn how it is doing the back end. Yeah. Yeah, is doing the back end. Yeah. Yeah, that's true, Matthew.

[2:27:44] Okay, we'll just wait uh two more minutes and then we'll begin the next phase where we have where we will be seeing what architect has done. We'll be attaching some knowledge base, some tools to the app that architect has

[2:27:58] built out for me. >> So yes, maybe if I can just add >> just just one thing because I guess people have been saying oh it's hard and people have been saying oh it's hard and it's it's difficult and stuff. So

[2:28:12] >> couple of things. One is see all of you obviously want credits uh this $20 credit that I think Su has given for all of you. You can use architect for free obviously a complex set of agents you might run into okay the credits get

[2:28:27] exhausted but most platforms whether you're going to use replet lovable claude there's a free version and a paid version right so I think you have to uh I mean some of these issues you'll run into which is I think just part of

[2:28:43] life I think that's one second is it seems complicated because you trying to learn like what is under the hood in terms of what does it need

[2:28:55] the hood in terms of what does it need to drive a car, right? Like you've seen most places it's like hey lovable shows you the car as the external piece but when you open the hood it's empty. We here essentially seeing each of the

[2:29:08] individual components come together which is which most engineers agentic engineers or for deployed engineers spending time dayto-day trying to figure that out. So uh it can feel overwhelming because it's 45 minutes. You spend you

[2:29:23] spend a couple of weeks on this. This is pretty straightforward. So uh most people that have joined us have been they come into a project they maybe join the rapid prototyping team. They build out some prototypes. They get good at

[2:29:35] understanding the nuance and then they go on to build complex uh solutions. So I think what you should take away from this is oh what are the agents under the Gmail? How do I connect with like a knowledge base?

[2:29:48] And then from there to what Suash was saying like pick a use case that's close to your heart right something that you resonate with and then from there kind of move and build something out you'll go deeper and you you'll not want to

[2:30:01] go deeper and you you'll not want to give up soon.

[2:30:15] architect has built out. I hope everyone is back. what we have done till now let me just uh give you an overview uh first in like

[2:30:31] uh give you an overview uh first in like 2 3 minutes. We had a prompt with us. We had a problem with us which we are trying to solve. We converted the problem into a prompt, a very structured prompt that okay I need this problem to

[2:30:45] be solved. Uh this is how the UIUX would look like. This is how the agents would look like. Uh this is the input. This is the output. This is how the flow would look like. We gave that prompt to architect in the homepage. After that

[2:30:59] architect asked me few questions. Uh if you scroll up the chat, you'll also be able to see all the things that we have done till now. It asked me few questions. I answered I gave it more context around what I'm

[2:31:12] trying to build and then it actually created a PR for me a product requirement documentation where it is itself thinking and defining that how the agent architecture would look like what agents it would need to

[2:31:27] look like what agents it would need to create what agents will have which model and what the temperature would be what the top P would be all the parameters of >> you need to share your screen I think. Just double check.

[2:31:41] >> Okay, let me just share it again. I hope it is visible now. >> Great. Yeah. So, here we are in the plan. So,

[2:31:53] we give it the prompt, we answer the questions, we get the plan and in the plan we saw what agents architect is going to create, what model and what parameters of the agents would be more relevant in this kind of use case. We

[2:32:07] got the user flow. We got the UIUX specification. We got the app mockup. We kind of change the app mockup ourself to modify it as per our need. We click on build agents after we got the app mockup. Now is what the

[2:32:24] real magic that happened behind the scenes which looked very simple. Uh let scenes which looked very simple. Uh let me just go more in depth around it. So here you see a flow of agents right and it might vary for uh other users because

[2:32:38] again as I mentioned uh LLMs are nondeterministic the end output will be same but the journey might be different so you most probably will have these so you most probably will have these three agents 99% of you so one agent if

[2:32:52] you click on this single agent you will see all the details of that agent and this is something which architect automatically has decided and created and this is very similar. I mean the UI is just different to what Anie was

[2:33:05] showing in Liser Studio. So how Lizer studio and architect are connected. Let me also uh give you a brief idea around that. Lizer studio is basically the base upon which architect is built out. Right? All the agents that architect is

[2:33:19] created it is actually created inside Ler Studio only. And you can actually redirect to let's say if you click on this open in Lizer Studio button you can see the same agent in the same LER studio platform which any was showing

[2:33:32] and the same agent role agent goal agent instructions will be represented here all the parameters all the top all the memory or the model which it is choosing it will be shown here and the same thing in architect itself we are showing so

[2:33:46] that you don't have to juggle between two platforms and if you carefully read the description engine role So let's take any one agent. I think we can take the research agent or the outreach agent itself. If you see the description that

[2:34:01] okay grounded is strictly on the per session seller knowledge base plus provided company research and a data selected persona drafts a personalized subject plus body and two three alternative hooks as JSON. This is a

[2:34:14] short description of what agent is doing. What role this agent would be following is something like expert B2B cold outreach copywriter who writes concise personalized emails grounded strictly in a seller's knowledge base.

[2:34:29] If you read the agent goal, it would be more in in detail that okay, I want this goal to be fulfilled by this agent that okay, draft a personalized cold outreach email and two three alternative oneline hooks for the selected prospect

[2:34:43] grounding all sellers claim strictly in per session knowledge base and all prospects company facts strictly in the party research. And this is the most important part of the agent where you are actually telling the agent what

[2:34:55] instructions to follow, how to work, what will be the input given to you, what steps you need to follow. So obviously role goal matters but agent instructions are the most important thing which agent would be uh having in

[2:35:09] context to make any flow or to perform any work. So this will be very detailed that okay you are a outreach agent and these are the inputs that you will be receiving these are the steps that we

[2:35:22] will be following and we can actually open in leisure studio and expand this agent instructions from here so that it is better visible so this is the agent and this is the task that it is performing based upon this instructions

[2:35:35] right inputs you receive is the target company is the user context steps that it needs to follow is that okay first of all resolve disambiguate the company. Uh second is gather an overview from the web. Third is find recent signals from

[2:35:49] roughly the last 12 months usable as outreach figures, funding, key hires, product launches, market expansion. So it will also take into consideration of the market and it will figure out that okay this is the best time to reach out

[2:36:04] to this particular person. Infer three to five business pain points tied to the company stage, industry and signals you found. Each endpoint has one lineal stay grounded. These are the output rules that okay how you should provide the

[2:36:17] that okay how you should provide the output. Uh so all this kind of is the instruction which we are given to the agent to tell how it should perform right and obviously we have selected perplexity sonar pro model which is the

[2:36:32] best model uh for this kind of research purpose. But if you open let's say a prospect agent right so this will be using enthropic clots on it right so depending upon the use case architect itself will decide which model to choose

[2:36:47] which uh parameters to set and how to set it what agent role should be there agent instruction should be there if you have to manually create this agent right without architect you have to come to studio you have to manually decide you

[2:37:02] have to first research that okay how should I give the agent instruction. How agent instructions? Which model should I choose? Which parameter should I uh enable disable? Uh should I enable memory or not? Should I attach any tool

[2:37:15] or not? So all these things are taken care automatically by architect right care automatically by architect right right now. And this is the agent which you can also use from an external service. So this is the agent API which

[2:37:28] I think Ani was showing in the previous uh session. and uh this agent JSON this agent inference API let's say if you're using some other software or some other internal tool or some other piece of software of your own you can actually

[2:37:43] service so you don't necessarily need to use this in lizer studio you don't necessarily need to use this in architect this agent can be used from anywhere so that is also one of the core functionality uh which we provide in

[2:37:56] leer studio if you think you want to use some other model of perplexity you can just come here you uh click on any other model. If you want to use Gemini or some other model, you can do that. Uh if you want to control the temperature top

[2:38:10] yourself, not rely on architect, you can do that. So you can manually customize the agents here. And also one other way is you tell architect to customize the agent for you. That is also possible. So let's say if you are feeling that the

[2:38:25] output that the app is giving you is not very deterministic or is not as per your uh as per what you want. You can actually tell architect that okay I want the language to be more concise and uh I want it to follow this three to four

[2:38:40] steps and ch it will change the agent accordingly. So every changes in the agent in the app in the UI in the UX can be done just by having a chat with architect. If you don't exactly know what you want to change and what change

[2:38:55] will uh have what impact on the app or on the agent right so this is the power behind of architect which it is doing behind the scenes in the front end it would look like three nodes right flowing around but these are three

[2:39:09] specific agents which we have created for a specific task that we are trying to perform and we will actually make it a little bit more complex when we give the second prompt of you know connecting the tool for connecting the Gmail tool

[2:39:21] for me. So first of all let's start with basics and then we'll move on to the more complex version also. So I think the agents part we have taken care till now if you guys have any doubts in the agents part uh in the yeah what is

[2:39:35] temperature? So temperature basically is a parameter. You can think of it like this. Temperature has value between zero and one. Right? And if you want the output to be a lot more deterministic, to be a lot more grounded on the

[2:39:48] instructions that you have provided, you can set the temperature as zero. Whatever you have asked the LM will only provide you that uh particular output. But when you increase the temperature towards one, you are increasing the

[2:40:00] creativity of the agent. the agent can basically uh have a lot of variety in you will never know what kind of uh response the agent is going to provide creative ways when the temperature is very high. So that is what the

[2:40:15] temperature uh does in a very simple terms if I have to explain is there ability to test different LM side by side to see which performs uh best cost speed and quality. I mean one way to do that is use our agent

[2:40:30] simulation engine which I'll be showing you in uh few minutes that is also where we can test out multiple LLMs and uh you can actually quickly if you want you can change the model from here and what you can do is we have a chat screen right so

[2:40:45] you can actually have a chat directly with the agent in leisure studio and ask it certain questions ask it to certain do certain task and you can in the real time see what the agent is performing behind the scenes and you can actually

[2:40:57] feel the difference between different models. are there building agents in VS code. We get a

[2:41:10] agent skill instructions file generated. Is there a same happening here or is there any difference? Uh so uh not exactly the same. The difference is how you define the agents on how you have the framework behind uh the agents.

[2:41:24] Right? So Lizer actually has its own framework of building out agents and building out the multi-agentic orchestration and everything. Uh and that will decide upon the use case that you're trying to build that okay uh how

[2:41:36] to build out multi-agent orchestration, how to build out a single agent but the agent itself. Yes, obviously there are like four to five parameters that defines an agent like you mentioned the prompt, the instruction, the role goal

[2:41:48] but different uh platforms let's say crewi is defining the agents in a different way. We have our own uh g agent which we are defining the agent in agent will always remain the same that will it will has to have certain skills.

[2:42:01] It will has to have certain prompt instructions role goal model. NLM should connect to a knowledge base. You can attach a tool to it. Can we use annotin? So anin is a completely different platform saga from

[2:42:17] lizer but we have a functionality called a superflow which is even better than an I would say and you can perform a lot more things in superflow which you can't do in any and superflow is actually a part of uh lizer studio. So you can

[2:42:32] actually go to leer studio explore superflow uh anytime you want to. Superflow is basically uh a way of achieving a very deterministic output where you can actually define the nodes of the agent that you are trying to

[2:42:46] build. You can actually define uh the LLM. You can create a workflow of agents which you want to do. So maybe we can actually I can actually show a few examples of uh super superflu at the end if you have time. But this is how the

[2:43:00] screen would look like. Uh it has a lot more capability than anything. You have a human in the loop. you have you can wait for approval. Uh so yeah this is also one of the feature that we have in Lizer studio.

[2:43:16] architect or does it cost? So as soon as you sign up as we mentioned you'll be having uh $20 worth of credits. Then when you use any agent or when you build consumed. Once those credits are exhausted then you will be uh um yeah

[2:43:32] then you will be actually need to do a top up and then get the credits back. So we have uh plans defined in architect and laser studio but obviously the first build the $20 will be able to cover the first build along with uh using the

[2:43:45] first build along with uh using the agents on architect and studio both. agents on architect and studio both. [snorts] able to build out the app and built out the agents if you can see the app that

[2:43:59] is built. Uh some people have already deployed I saw in the chat. Uh so maybe deployment also I'll just tell you how to do it

[2:44:18] given in the super flow. If you check out uh prau uh you can try that examples or you can try building out something for yourself also if you have any deterministic workflow in your mind.

[2:44:38] an app built out inside architect, right? Uh so this is the final app that we will be building but this is what we were building now. So okay whenever you have built out an app there are like few other options which you can also explore

[2:44:53] uh with the app that you have built. We can actually click on this preview button if you need to open the entire app in a full screen mode so that you can see how it would actually look like uh in the

[2:45:07] full screen. And here you can actually use the app in a much more better way. We have a option to deploy which I'll be talking about in few minutes. We also have a option to connect your GitHub. So what if I want all the code of architect

[2:45:25] uh with me uh locally or residing inside my GitHub account. So I can actually connect my GitHub account. I can take the entire code which architect has written and I can actually work on top of that code locally. Let's say if

[2:45:38] you're using cloud code or codex or some other thing to work upon, you can basically build on top of what architect has built and you can push back the code architect. So that is also some functionalities which we give to uh

[2:45:51] developers or if you have some technical uh member in your team who you can work with. This is also something which we provide. Now deployment is one thing. Uh it's the final step in building app. So let's say if you have built this app, if

[2:46:08] you have done all your changes in the app, if you think your app is final, at the end you can actually deploy the app or even if you deploy it now, it won't be an issue. You can redeploy the app. But the ideal way to is to uh deploy the

[2:46:20] entire app after you have done all your changes. Now let's actually see the preview. Right? Once you click on this preview button, uh the full screen mode will open and what it has built out for me. First of all, I have to upload a

[2:46:34] knowledge base about the company doc, right? It is saying to me that upload your company document. So again, I am sharing this link in the chat from where you can download the knowledgebased document. So you don't have to create

[2:46:48] one. Once you click on this link, if you see in the left sidebar, there is a knowledgebased section. If you click on this knowledgebased section, you can this knowledgebased section, you can just click on this download. docx button

[2:47:01] document for you the knowledge base document. So for example we are actually using lizer company knowledge base here and if you uh open this document you will actually see uh all the details that okay what lizer does the product

[2:47:16] proposition what is our ideal customer profile uh something related to pricing oneline positioning so I have basically created a simple document of what lizer

[2:47:28] is what lizer provides you how it will help enterprises how what is our value proposition and if You click on this docx just save this file uh locally in your PC. So we'll be needing this to upload in the knowledge base. This is

[2:47:43] the file that you guys should be having. Uh let me know if you are able to Uh let me know if you are able to download the file. sandbox uh wasn't found. So errors will be seeing uh uh an uh in few minutes.

[2:47:58] But the best way to solve errors is to click a screenshot of the error and paste the same error in the architect chat and ask it to fix it. Uh that will solve most of the errors. Architect has the capability to solve any errors

[2:48:12] itself. Also there is a fresh preview button here. So let's say if you are not able to see the preview or if you face any issues while loading the preview, you can actually click on this refresh preview button and it will reload the

[2:48:25] preview for you. Or you can just directly click on this uh preview button directly click on this uh preview button to work in a much more comfortable way. But uh is everyone do able to download the document which because we'll be

[2:48:38] needing that document to upload and to make the to test our app. [snorts]

[2:48:50] are able to I'll just share the steps again. I'll share the URL. will see a page like this and in the page left side there will be a knowledge

[2:49:06] base section. You can click on this knowledge base and you will see a small section like this where we are giving the riser company KB document. You can click on download.d doc. Once you click on download, a docx file will be

[2:49:24] on download, a docx file will be generated and saved locally in your PC. Just save it for now. How when should I upload? Yeah, we'll just go through it. upload? Yeah, we'll just go through it. Let everyone just uh save it.

[2:49:39] We are just saving it and then I'll be explaining you how to upload it. in progress. uh but there is no output. So two ways to solve this. Uh one way is

[2:49:53] that okay you click on that refresh preview button. If that is not working you refresh the entire tab. If that is not working you give the same prompt to architect chat. So whatever issues with architect that you are facing as I

[2:50:07] mentioned architect itself is able to solve the issue. So for me no issues came so I'm sorted. But in case you are facing any issues, you can actually paste the same issue or error here which you are pasting in the chat and ask

[2:50:21] architect to fix it and hit enter. That is the best way to solve any issues. Uh most of it will be preview related that you can just kind of try to refresh the preview or refresh the entire chrome tab that will also work.

[2:50:40] where I upload this file? as we'll be telling you in few moments. the document with you once you see the app once you have the app open a preview

[2:50:55] in a bigger full screen you just need to uh upload your document in this particular section where it should be mentioning something like knowledge base your company and if you can just click on uh the path where the file was saved

[2:51:09] you can click on open and this knowledge base will be ingested uh inside this base will be ingested uh inside this app. So just follow this proper step to upload the knowledge base. Let me know if you're facing any issues.

[2:51:32] upload the docx. Where is the preview but preview button is the top right? top right you will see a preview button. Here you can click to open the app in a full screen mode.

[2:51:53] that is not what I downloaded. Uh you should be downloading this uh in the URL the chat there will be a knowledge base section and this download.d do6 button you should be clicking section two knowledge base. This is the document uh

[2:52:08] that we need to upload once the app is built out here.

[2:52:24] the document, uh architect now can refer to this document while creating any content or while generating any uh outcome for you. Right? So [snorts] let's say we want to uh we have a target company which we need to pitch our

[2:52:39] product or which we need to uh sell our product or which we need to outreach right. So we can actually paste any name of the company which is relevant for our product. So let's say Accenture is one of the company which is relevant for

[2:52:53] of the company which is relevant for architect. So I'll just share the architect. So I'll just share the company name also in the chat. Sir, I don't have a doc in the KWL base on my laptop. Uh so you will need to

[2:53:07] download the documented document that we have uh given in the URL and then you need to have the document in the local laptop. Uh then only you will be able to upload that I have downloaded right file but in my

[2:53:20] app your organization section have different doc uploaded already. Uh it's fine. So for people who have already certain kind of uh document attached uh in the app that also that will also work. So what architect does sometimes

[2:53:34] is it will also create dummy data for you u that okay this is the app and this is how the UI would look like. So for some of the users you will already be seeing some of the data being uh uh generated in the app itself. Let me give

[2:53:47] you an example. Uh okay we don't have an example here for some of you guys you might also see some uh prefilled text or pre-filled data already present and uh that is also fine that will also work you will be

[2:54:02] having a toggle button somewhere on the top right to you know toggle on of that dummy data which you can use to toggle off the dummy data it will kind of remove the dummy data for you and you can actually uh put your own content in

[2:54:14] can actually uh put your own content in it. company uh filled in the company name, you can actually click on research company. And while you are clicking on research company, you can actually see

[2:54:30] what are the agents that is powering this app and what are the agent that is working currently. So the dot which is kind of glowing and dimming is the agent would look different for different users. So it will not always be same. uh

[2:54:47] this section might come in a left sidebar. So uh it will generate different value but the end outcome we should be able to uh generate we should be able to achieve right and I have just given it uh the name of the company I

[2:55:02] have given it the knowledge base and I have click on research this company right so what is happening behind the scenes I mean obviously I can see all the lot of content getting generated automatically right but how is this

[2:55:14] working so let's go back to architect app what is happening behind the scenes is whenever I am clicking on research company an agent is getting triggered right if you go to agents tab we have a research agent right whenever I am

[2:55:30] clicking on research company this research company will take Accenture as an input will take the knowledge base as an input and it will follow the instructions given to this research company that okay you are a research uh

[2:55:43] agent the inputs you will receive is the target company Accenture which we have given the user context is the knowledge base that we have given And then it will follow these steps right it will follow these steps and it will generate this

[2:55:56] output for me critical output rules is there uh what steps needs it needs to uh do. So what we have done behind the scenes is we have used this agent. Let scenes is we have used this agent. Let me also show you the agent in the lizer

[2:56:10] studio much more detailed way. These were the agent instructions which we went through few minutes ago. Right? And based upon the inputs, this agent has worked and generated an output. And that output I'm able to see in my final

[2:56:23] application which I have built out which is coming here. And you see this is like a proper structured output which I can see the research summary for Accenture right I can see the overview that okay Accenture is a global leading

[2:56:36] professional service company. The recent signals what is happening in June 2026 Accenture expanded its partnership with major cloud providers. inferred pain points what are the problems Accenture is currently facing all this information

[2:56:50] I will need to generate my outreach personalized outreach content that's why the first step was always to figure out about the company and the second step prospects inside the company and the third step is to generate the outreach

[2:57:05] content and the fourth step which we are going to do sometime after is sending automatic uh email to that particular prospect right so it has also researched about the pain points that particular Accenture companies facing.

[2:57:19] Talent retention and up scale up scale scale up scaling is one of the problem that it is maintaining a workforce of over 700K while ensuring proficiency in Genai is a constant operational challenge for them. Scaling generative

[2:57:33] way from pilot to production is a bigger pain point and this painoint is also that is what we do. We deploy your agents into production. We help you make production ready agents. Uh they're also

[2:57:46] facing issues managing complexity in global cloud integrations. Also they points right but since we have updated our knowledge base it has context of what proposition Lizer is proposing what problems Lizer is solving. So it has

[2:58:00] automatically kind of showing me the pain points which is more relevant for pain points which is more relevant for me uh which I can help as a company as lizer for this particular uh accenture in their helping them with the painoint

[2:58:12] in their helping them with the painoint side. I can actually uh see some of the suggested prospects whom I can reach out to. Uh there is a global lead AI and digital transformation. Uh there is a VP managing director of cyber security and

[2:58:26] manag services. So uh I can actually select any one of the prospects from here or let's say if I have already have a prospect I can actually put the name and uh the role of the particular person and search about that particular person

[2:58:39] also. Once you have selected a prospect that okay this is the more relevant targeting to uh let's say I want to target a global lead in AI and data practice right I can actually generate

[2:58:52] personalized content for that particular person. I have the detail uh information about the uh department he's working in. I have a detailed information about the entire company, entire organization, what pain points they have facing and

[2:59:06] a document in the knowledge base that we have already given. So when I have all this information, I can generate a very detailed uh outreach content for you.

[2:59:18] Right? And in case you encounter any issues in generating outreach content showing how to fix issues right. So when I click on generating out outreach which is happening inside architect right something the agent is not able to

[2:59:33] do properly which we need to figure out. So there are two ways I can go to leer studio I can do try to manually figure out what is happening what went wrong uh why is it not able to work or I can just take a screenshot of the error that I'm

[2:59:45] seeing. So I can take a screenshot of whatever error seeing and some of you might be getting an error, some of you might not be getting an error. Whoever is getting an error can take a screenshot. Come to architect the

[2:59:58] architect app that we were building and just paste that error. Paste that error screenshot in the tab here or you can click on this plus button and uh add assets and then add the attach the image file there. And what I would be doing is

[3:00:14] I would just be asking architect that okay I am getting this error while generating outreach content. outreach content. Can you fix this?

[3:00:34] of you know trying to solve problems whatever you are facing automatically without uh applying too much technical knowledge or without even thinking about what I need to do what I need to check obviously all those functionalities are

[3:00:47] check the traces where the agent is failing uh what was the issue you can solve it manually but architect gives you the capability to solve any agents related to the agent or any issues related to the UI

[3:01:01] And it can also it will also tell you what was the issue what it is trying to solve how it is trying to solve. So if you are uh curious enough you can read the entire chat of how is it trying to fix the issue which you are facing right

[3:01:14] fix the issue which you are facing right and after it has done solving the issue we can try again the app. So let me know if you guys are able to generate the content, if you are able to generate the prospect details, if you are able to

[3:01:29] prospect details, if you are able to generate the uh rearch of the company, let me know at which phase you are currently in so that I'll be able to help you in a much more better way. Edit agents is definitely a cool feature.

[3:01:41] Yeah, that I have not seen anywhere. That's true. How do I remove the one I attached and add another? So Ranjan there is one very simple way to do this. You ask architect that okay I have a document with me. I want to attach that

[3:01:56] document with me. I want to attach that document in the uh UI in the app. I don't want you to create a dummy data or a dummy document for me. So uh all the are facing while building or that you are facing or that you're trying to do

[3:02:11] writing here that all the queries if you write to architect it will actually do and solve the queries for you. Uh that is the uh biggest feature I would say that everything is customizable everything is changeable just by asking

[3:02:27] architect to do it. Um if you are seeing any dummy data and you want you don't want the dummy data if a document is already attached and you want to remove that document you just ask architect to remove that document. Uh if you don't

[3:02:39] like the UI if you would like a much more uh complex UI if you would like a sidebar there you can ask architect to do it. If you don't like the want to do some changes in the agents

[3:02:54] you can ask architect to do that. So everything is completely customizable and uh completely iteratable just by giving it a prompt. Uh no coding or no uh manually changing of agent instructions role goal models you will

[3:03:08] need to do. So let me know if you are uh uh able to you know get to reach to this phase where you are able to see the app where you are able to generate some content out of the agent

[3:03:22] or if you are able to face if you are facing any issues is architect able to solve it for you once you have given it as a prompt.

[3:03:54] able to generate the outreach content also for me. So what basically happened right? I faced an error in architect with the agent that okay the agent was throwing me some error. I didn't even check what the error was and I didn't

[3:04:07] even go to the agent and uh think about what the error is and how the error was coming what I need to do to solve it. I just take a screenshot of the error. I just went to architect. I asked architect in a very simple English that

[3:04:22] okay I am getting this error while generating outreach content. Can you fix this? Right? And architect actually fixed this error for me. And what was that specifically the technical error if someone wants to

[3:04:35] chat that okay this was the issue architect fixed it in this particular way and uh for someone who just wants the issue to be fixed you can directly come to your app and have a chat with the app again or try using the app again

[3:04:51] and check if that is able to generate the content for you or not. And if you see I have a drafted email ready for me for the global head of AI chief AI officer of Accenture. So I have the subject with me Accenture Jenni

[3:05:06] expansion a faster path to shipping industry specific agents. I also have industry specific agents. I also have the body with me the body of the email. So it is also taking the context of what is happening around the uh news right.

[3:05:19] So it is first uh writing up a very personalized content that okay congratulations on Accenture's June announcement expanding your ji practice into supply chain manufacturing uh building industry specific agents that

[3:05:32] complexity compounds fast that's precisely the problem lizer is built for uh lizer is an agentic it is pitching about lizer now leer is an agentic mid-market inter companies that are actively building a into their product

[3:05:47] given your team's push to certify 90% % of the workforce in AI governance and enterprisegrade agent building foundation would take real pleasure of your delivery leads. Would it be worth 20 minutes call to explore whether Lizer

[3:06:00] could accelerate what your practice is already uh building? Happy to work around your schedule. Right. So where this content is coming from first of all it is a very nice personalized outreach email uh that I can send to uh someone

[3:06:13] like a global head of AI right in Accenture automatically. This content is coming from two three things right. So first of all we have given the agent all first of all we have given the agent all the information about how the

[3:06:26] uh how to how to propo how to give someone uh detailed overview of what Lizer is doing in the document we have written what is our value proposition uh what are our products that will help you to solve this particular problem. Then

[3:06:41] one of the agent actually did research that what problems my target company is facing, what are their pain points. One agent actually figured out that okay whom should I reach out to in this particular company and based upon all

[3:06:53] these details that we already have the uh company details, the prospect details, the knowledge base uh in which I have provided all the details about I have provided all the details about Lizer. It actually generated a very nice

[3:07:06] personalized outreach email for me, right? And it is also giving me an I just mentioned that can you give me three alternative hooks which I can choose from. So it has also given me three alternative hooks to uh kind of uh

[3:07:19] include in the mail include in the mail body or change in the mail body. And it is not only limited to this. Let's say if you want to go one step ahead if you want to add more functionalities to this app that is also something which we can

[3:07:32] do just by prompting which we are going to see uh in few minutes. But uh let us know let me know if you are able to generate the outreach content till now if you are facing any issues. Sir, we can can we use cursor to fix the

[3:07:46] generate the code index to fix the product development. Yes. So actually if you are a developer you can take the entire code of architect in GitHub. You can actually work on that particular code uh locally using cursor or using

[3:08:00] clot code and you can push back the entire code to architect back again. entire code to architect back again. That is one of the feature that we have.

[3:08:17] is able to you know kind of get to the end part where they are able to I was able to resolve multiple errors and get imaginated. Wow. Great. Great.

[3:08:30] facing that everything is uh resolvable. So I would just say if you are in need of uh any error how is [snorts] your difficult to debug the code. So actually you don't even need to debug the code. If you are a normal user and you don't

[3:08:44] have any technical knowledge you just need to tell architect the problem that around the problem. If you share a screenshot it will be much better. If it in the architect chat. So we actually fixed one problem ourselves site. I

[3:08:58] didn't debug the code myself that what was the error. uh I just tell architect that okay I am facing this error fix it for me. If you are a developer you can You can try to debug it yourself or using cloud code or using cursor however

[3:09:12] you like. [snorts] Okay. So if you go back to your app there are two ways.

[3:09:25] You can connect your GitHub from this small GitHub icon that you are seeing and take the code from there. You can clone the repository which architect is pushing into GitHub. Otherwise, a short way is to just click on this three dots

[3:09:38] It will give you the zip file of all the code that architect has created for the code that architect has created for the app and you can work on it locally.

[3:09:56] science student is very useful for me. Yes. Uh Aira actually you can build out uh pretty you know uh useful use cases uh related to whatever problems that you are facing uh around you using architectural using leisure studio

[3:10:13] email not generated email is the second part which we are going to do now uh which we are going to try now actually let's we can actually go to that so what if I want to take this one step further and I actually want to connect my Gmail

[3:10:28] to this part to this particular app and I want the email to be created as a draft in my Gmail right so that I can automatically send that mail to someone which I want to so again the same thing can be done just by giving architect a

[3:10:45] prompt and for this uh if you don't even need to uh try prompting it what we have done is again the same document which I've been sharing multiple times if If

[3:10:57] you click on this document, you will have a build prompt and you will have a connect Gmail. Right? The fourth section in the connect Gmail, you will see a iteration prompt which you can directly copy it from here and it's nothing uh

[3:11:12] special. It's just a prompt saying can you also add a agent that can connect to my Gmail and create and create a mail draft in my Gmail with the generated outreach content upon clicking save content as draft. So in very simple

[3:11:25] plain text I'm telling it what I want right actually I will just copy this prompt and paste it in the chat so that you guys can copy directly from here and once you have this prompt you can directly come to architect and in the

[3:11:39] directly come to architect and in the bottom left chat box that you are seeing you can paste this prompt and hit enter and architect will start building the agent which can actually connect with Gmail which can send the Gmail uh which

[3:11:54] your Gmail so that you can automatically uh send this email to anyone which you would would like to. Uh this is also by the way same kind of use case which you can build out in architect is like if

[3:12:06] you want to reach out to multiple recruiters like right so you can give your resume to it in a knowledge base. You can actually give your uh CV, you letter or ask architect to generate a cover letter based upon the resume. You

[3:12:21] out the companies that you want to apply with as per your interest as per your uh uh domain and then ask architect to create outreach content for them that will send out the CV and that will send out the uh resume to them automatically

[3:12:35] over mail to recruiters. So similar kind of use case uh which we are which you have built now this can be replicated along uh a lot of pain points that you are guys are facing in your day-to-day life.

[3:12:48] Yeah. So everyone who are here till now I'll just paste this in the chat again and hit enter. So that we can also attach a tool to the agent and can try and can test the Gmail functionality

[3:13:03] also. How to connect a tool to a particular agent. particular agent. Let me share the prompt here. Sir, if I build an agent conversational agent? Yes actually a lot of use cases

[3:13:16] need a conversation conversational kind of UI also. So if I ask architect to in the same app also if I can ask that give me a chat interface where I can chat perform certain action or do certain kind of research analysis that is also

[3:13:31] something which it will be able to do. So a chat agent will be created. So can connect and generate the outreach draft and connect Gmail tool uh to the that okay, I need a chat interface, a chat agent, it can also do that.

[3:13:46] Can I use in my own built website deployed in versel? Uh, yes, that is also possible. So, you will need to uh download the code from for that particular project that you have already built out and paste that file as a zip

[3:14:00] file in architecture and then ask it to build on top of it. So, you can do that also if you have already an existing project that you need to work upon. Yeah. So try pasting this prompt in architect uh and it should be able to do

[3:14:15] the work. Uh it will take four to 5 minutes to create another agent which could connect Gmail for me. Uh try it out and let me just uh bring out my charger in a second till this is building out because my battery is about

[3:14:29] building out because my battery is about to die.

[3:15:08] building. Uh, let's see if there are any questions let's see if there are any questions till then.

[3:15:22] I mean, uh, we'll be sharing the coupon code with you guys. So you'll be able to get it. Uh meanwhile I would just recommend try opening architect or try opening laser studio in an incognito mode

[3:15:37] because there might be some cache already in the browser which we need to already in the browser which we need to clear.

[3:16:14] months. So it is unlimited actually sol you uh your credit will not expire once you uh your credit will not expire once you have purchased

[3:17:00] Okay, it has actually generated the agent for me now.

[3:17:14] create the tool, right? And if you scroll down at the end you will see all the chats that architect is following

[3:17:26] created a Gmail draft agent. It is now testing the Gmail draft agent correctly and is telling you that okay what I have

[3:17:42] and is telling you that okay what I have built a Gmail draft agent a newer agent connected to your Gmail via the Gmail integration. So let me know if you guys are able to do this.

[3:18:12] the test data the company detail disappear. Uh yes. So Subra we in the beginning if you remember let me show you uh in the beginning there was few questions right whether you want to attach a database to it or

[3:18:26] not. So we selected no for this particular use case because we were trying a oneshot kind of a iteration. We are trying a generate the content in in a short but if you need to save the data in the starting there is was a question

[3:18:40] will this app need to save and manage data across sessions. Right? If you have database automatically for you. And if the data. So it won't get refreshed or it won't get it won't get removed in

[3:18:53] it won't get it won't get removed in every refresh that you're doing. So the guys who are able to connect to the Gmail tool, uh let me know.

[3:19:15] And once you have connected the Gmail tool, you can try the app again. Upload the knowledge base. And if you have already connected a database, then you won't need to fill this details in again and again.

[3:19:28] But we are doing it in a one short way. So let's just do this. generate the prospects and then we'll just see the last step of connecting the

[3:19:42] just see the last step of connecting the tool. if you can select a prospect, you can generate the outreach.

[3:20:02] for people who have been following along, one small step that you need to do to connect your email is to basically authenticate the tool. Right? If you guys are seeing this screen and if you guys can see a pop-up like pending

[3:20:15] integrations, pending tool integrations, you can actually click on connect now and you can click on this connect button. If you click on this connect button, you will get a authentication system like this authentication page

[3:20:27] like this where you can just select your email ID which you want the agent to connect to. your Gmail and once you have given it access to connect it to your Gmail, the agent will automatically get connected

[3:20:39] to Gmail and you will see a all done kind of a green signal here. this prompt and if you have connected the Gmail tool. It's pretty simple. Uh

[3:20:55] there is also in the agent section you will see four agents now instead of three agents. The first is research agents, prospect agent, outreach agent and the fourth is Gmail draft agent. If you click on this

[3:21:07] small icon you can also again get the Gmail authentication popup where you can connect the tool. You'll just need to authenticate your Gmail account for it.

[3:21:22] it sir. This Lizer studio has a lot more features compared with emergent. Yeah, emergent has a very uh basic feature list. I would say Lizer studio and architect will be having a lot more kind

[3:21:34] architect will be having a lot more kind of a uh feature list with us and we are of a uh feature list with us and we are always more enterprise ready. always more enterprise ready. Uh okay, I think people are trying out

[3:21:46] connected the Gmail to the agent and once you have given the authentication you can actually click on this agent open it in Lizer studio you will see that the Gmail tool has connected here in a detailed way that what actions this

[3:22:00] agent can take. So this agent can actually create email draft. This is the action which I can see in the tool configuration right. So which means that this agent can also create a draft of email in my particular email right which

[3:22:15] I have connected it to in my particular Google account to the app which I have built out. So if you can click on the app preview you

[3:22:27] will also get the option to send or to generate the email in my Gmail account. So I just need to follow the steps of uh attaching the knowledge base giving it attaching the knowledge base giving it the target company

[3:22:41] researching and in the last step which was not before we will just get one more or connecting the Gmail which you have already done and after that step you'll be seeing a nice input box in the UI where you will be able to see a button

[3:22:54] to send or to create draft email for me. So let me just show you that okay it is loaded now let me just click on any one of the prospect and click on

[3:23:09] on any one of the prospect and click on generate outreach meanwhile it is drafting I can actually show you how it would look like. So the

[3:23:26] show you how it would look like. So the same flow which we have done till now whatever app that you have built in just the knowledge base uh do a company the knowledge base uh do a company research

[3:23:41] send this to a particular recipient or if you want architect to uh kind of automatically fill this recipient field for you. Uh let's say for now I want to see this email in my account only. I can actually uh click on this recipient

[3:23:56] email uh as my email id my personal email id which you guys are using and then I can click on save content as draft. So why I'm saving it as draft and not automatically sending is because I also want a human in the loop which I

[3:24:08] first need to review the email myself and then only send it. So it will which I can review and then send manually. I can also kind of do this confidence of what architect has generated for me and I'll just click on

[3:24:24] generated for me and I'll just click on saving draft to email. So uh let me know if you guys are able to follow the process till here. So replying to is it a desktop or web app? Correct. It will be a web application.

[3:24:37] Uh yeah my pal can you just repeat your question sir just repeat to connect Gmail and show me again. Yeah, we can go to the steps very quickly once more. So,

[3:24:49] whatever agents that you have built, if you click on the agent section, you will see one specific Gmail agent, Gmail draft agent attached to it now, right? If you click on this Gmail icon, if you click on this Gmail icon, you will see a

[3:25:04] connected. That's why I'm seeing this connect button. You will get a connect button here. Once you click on that connect button and if you loging in with your Google account, you'll be able to connect to the Gmail.

[3:25:21] Gmail icon and then connect it from there. That is the flow. So Gmail not showing. Uh if the Gmail icon is not showing, you can actually ask

[3:25:36] architect I want to connect to my Gmail. can you please uh connect my Gmail to it? It will all it will attach that Gmail uh icon to that agent or you can actually use this particular prompt which we shared earlier again so

[3:25:49] prompt which we shared earlier again so that it would be able to create that. you guys. If you have given this prompt to architect then only it will work which we just you know

[3:26:07] specifically tell architect what you want to do that you are getting again we can just ask architect. You can tell architect connect my Gmail. It will automatically attach the tool to the agent for you

[3:26:29] the option of uploading a file is disappear. Okay. have built, if you are able to uh get the content, if you are able to uh

[3:26:44] generate the outreach content or whatever you are able to do, uh after this just uh again I'm reiterating try to enter this prompt which I just pasted in the chat. It will attach a Gmail to your particular agent. And then once the

[3:26:59] Gmail is attached, you just need to click on the agent tab and you will see a additional agent added which will be named something like Gmail draft agent. You can click on that Gmail icon and then click on the connect button.

[3:27:24] this. So this is like one more app similar kind of app. So the same flow are trying to build the app might look different but the process will always be the same. Uh once the outreach content is generated you can add a recipient for

[3:27:38] email to anyone. I am sending it to myself only. So I am writing my recipient as my email id itself. And then I am clicking on this save content as draft button. You might be getting some other kind of uh text in the

[3:27:54] button. So the text might vary but the flow would always be the same. Once you click on save content as draft whatever email that architect has created it will kind of save the same email draft in my particular Gmail which I have connected.

[3:28:09] So let me also show you how it would look like in Gmail. So once let it save first and then we'll see uh once you click on save content as draft uh and if you see that draft save to mail the mail or the email outreach content that

[3:28:24] architect has generated will also reflect in the Gmail uh once you log in with the same account in the draft section if you refresh you will see so this was from the first app the which we built and this is from the second app

[3:28:39] which we just tried out. So these mails should automatically be saved as draft in your Gmail draft section. If you click on this particular mail, you will see the same body, the same content and I can just verify this mail myself. Uh

[3:28:53] uh modify something if I need to. I can actually modify it from the architect itself. Uh and then once I'm sure that okay this mail looks fine to me, I can okay this mail looks fine to me, I can actually send this mail. And I purposely

[3:29:05] because I want to make sure the mail which I'm sending is correct. I can actually automate the entire flow myself using architect that okay send this mail automatically even without my permission or uh it won't require a human human in

[3:29:19] the loop in the process right so it can also do that the only difference at that part will be so right now if you click on the Gmail out agent it has the capability to send the draft of the email right to create a draft of the

[3:29:33] emails if you go down and if you see the Gmail create email draft action it can perform we can actually edit this action and give them give this action to send mails and once I have given this the capability to automatically send mails

[3:29:46] capability to automatically send mails it can also uh do that and that also I don't need to come to studio and kind of manually attach a agent action or manually attach any different tool I can just come to architect and ask it I

[3:30:00] don't want a draft of the agent you to send the mail automatically So this capability or whatever architect or you want to add in the agent or you want to add in the app

[3:30:15] everything is just possible just by asking architect if you're facing any issues if you're facing any uh functionality UI problem if you are uh ready to add more functionalities more features everything can be doable just

[3:30:29] features everything can be doable just by prompting architect to do it. So we are a little short of time. So I'll just uh let me know if you guys are able to do this, if you're able to reach here. Uh so I know it's a lot in a brief

[3:30:43] multiple things. We have created the agents. Uh we have uh built out the UI. First of all, we have built out the PRD. We have connected a tool. We have sent

[3:30:57] email to the prospects. We have geted we the mail in the draft section. So a lot of things have been uh going on behind the scenes which is making it to work. connecting even though I have received a notice from Gmail to access any

[3:31:11] recommendations to solve uh Gmail access. If you're not able to solve I would say one way is to kind of first of all disconnect any tool. So ask architect that okay can you disconnect and reattach the Gmail tool in my agent.

[3:31:24] So I will also write down the prompt for you. Can you reconnect? Can you disconnect the Gmail tool and attach to the agent again? If you can give this prompt, architect will disconnect the tool for you and

[3:31:39] then reconnect it and then once again you can try to uh authenticate with the you can try to uh authenticate with the Gmail

[3:31:51] whatever error that you are saying or seeing or whatever problems that you are facing just give this same error to architect and 99% of the cases it will architect and 99% of the cases it will be able to solve it for you because we

[3:32:04] saw right we got got a one very good use case where we were facing some error and I asked architect to fix it for me and it was able to do it. So every time 100% of the things will not work in the first iteration because behind the scenes as I

[3:32:17] mentioned again and again that an L&M is working but uh always if you tell LM specifically that this is the problem that I'm trying to face and how you can um how you can help me fix it and also just instruct me to fix it give it more

[3:32:31] it a screenshot of the error and it should be able to do that itself.

[3:32:45] also a good thing. Uh I can't upload doc. I would say just prompt architect document. Can you fix that for me? Can you help me with that? And it should be able to do it.

[3:33:03] users uh for with Gmail connection is not showing. So I am sharing the prompt again which will help you connect the Gmail tool in the agent. So you can just paste this prompt in architect.

[3:33:24] My website is make in anthropic. Uh what do you mean by this? uh Shri so we are using a cloud model I mean behind the scenes to generate the content but the actually working one was we are using solar perplexity for research doing the

[3:33:41] research uh we were using uh claw to generate the content for you but the entire app is built inside lizer itself uh the architect which we have just been using and also whenever let's say You are done with the app, right? You are

[3:33:59] are done with connecting the tools and everything. What ideally you should do is now you should be able to deploy your app. What the deploy button will do is it will give you a deployed URL which you can share with anyone which you can

[3:34:14] show to everyone that okay this is what I have built it in uh few hours or few I have built it in uh few hours or few um minutes. Right? So I have I am pretty much satisfied with what I have built. I can just click on deploy

[3:34:29] able to reach still the Gmail functionality I would say still you should be able to you should reply deploy your app because whatever built the three agents the creating the outreach content the doing the research

[3:34:42] for the uh companies right that is also a good section which you can show it to the world right and uh you can click on this deploy button if you want to push this particular app to a marketplace so architect also has its own marketplace

[3:34:56] where you can showcase what you have built and uh whatever you are working otherwise if you want you can also disable this. So if you're publishing this to marketplace I'll also show you how it will look like.

[3:35:16] section right in this marketplace your app will also be featured once you have built it out and you have enabled that toggle and you have deployed it. So these are all the built apps by users which anyone can come and see that what

[3:35:28] people are trying to build and use it. But in case if you want to uh publish you can just have to add a line or two for the description of the app you can generate the AI generate the content using an AI also and just click on

[3:35:42] deploy and publish to marketplace. For those who don't want to publish it to marketplace you can keep this toggle off and then deploy. for those who actually want to use their custom domain. Uh you can actually add your own custom domain

[3:35:55] to your particular app because right now when I deployed this I'll get a URL like this which is a architect.space URL but let's say if you want some abcd.com or some.com or somein domain you can actually purchase that and then use that

[3:36:09] particular domain to showcase your app. So whatever you have built I mean whatever uh part till now you have you are able to achieve I would say you should deploy your app first you can also keep on building after deployment

[3:36:22] so let's say if you're not able to connect uh Gmail to now you can deploy rebuilding you can click you can try attaching the Gmail tool you can add features and then once you are done you can just click on redeploy once you have

[3:36:36] can just click on redeploy once you have added new changes on your app and till this deployed URL. Once you click on this deployed URL, it will take few seconds to load in the first click only. After that this URL will be available

[3:36:49] where app will be completely hosted and you can share this URL with anyone. Anyone should be able to access your app. So let it load. Meanwhile, let just app. So let it load. Meanwhile, let just see if anyone is able to deploy the URL.

[3:37:02] If anyone is facing any issues sir whether we can make any agent to take data from WhatsApp messages. Yes. So all the tools that you want to connect to the agent right you can actually go to laser studio and there is

[3:37:15] a section of custom tools. If you click on custom tools you can add any of the external tool that you want to connect the agent to. So let me also show you if you click on connections in leisure studio and tools there are by default a

[3:37:30] lot of tools that we already provide you right if you can see notion Gmail uh brave search team slack YouTube if the your tool Instagram if your tool is not listed here you can actually request a tool uh or you can actually use a custom

[3:37:45] your WhatsApp APIs and everything ready to use here so that you can authenticate with your WhatsApp and then use that tool. So that is also part of the tool. So that is also part of the functionality in Lisa Studio.

[3:38:03] just give me a thumbs up of guys who are able to at least deploy your app. uh even if you have built out certain part of the app uh we have covered a lot of of the app uh we have covered a lot of it in the short time that we had but if

[3:38:16] are able to you know at least generate the outreach content or I see a lot of thumbs up great um so that is deployment first time it will take uh its sweet time but uh after it is deployed once so

[3:38:32] if you can refresh or if you can see my screen this value is now deployed I can actually copy this URL and share with anyone the deployed URL. Let's open it in incognito. Then it won't take any time. It will load instantly. So this

[3:38:47] deployment time will take its sweet time only during the first iteration.

[3:39:00] URL. This is not a deployed URL by the way Shri which you have shared. Uh the URL which you see here is the app URL to open the architect app. Deployed URL. You have to click on this deploy button and this URL will be the deployed URL.

[3:39:13] and this URL will be the deployed URL. You can copy it from here. deploy your app so that you know you can actually at least share uh whatever you

[3:39:27] actually at least share uh whatever you have built with anyone. uh if you were able to build this use case manually using any other uh framework like lang chain lang graph it will at it would actually take you at least few hours

[3:39:41] just to create a single agent uh and that part we are able to achieve in a matter of two hours I would say to create four agents and one agent was taking a knowledge base one agent was connecting your Gmail one agent was

[3:39:55] generating the content one agent was doing the outreach and at the And you don't even need to take care of the UI part. Right? So everything is doable in this particular short amount of time. Uh if you are able to build this uh same

[3:40:08] app using any other uh manual way of coding or using any other framework, it coding or using any other framework, it would take a lot of time for you guys.

[3:40:22] deployed their apps or even if you're not able to achieve any certain section of the program that we have uh tried here. So in connecting tool generating outreach content try it out uh after the seminar

[3:40:38] also I would say and make it work. Most of the most of the errors that you are itself. And for people who have already deployed it, for people who have already deploy it. And for the people who have deployed, I would recommend you to share

[3:40:54] your uh deployed URL in LinkedIn or in Twitter tager tag #simplearn. And we actually have a very good uh reward system of all the

[3:41:06] top people who will be getting a heavy engagement on their post. So we'll choose top three to five guys uh who have done LinkedIn post and who have deployed URL. If you can share a screen recording of how your app is working

[3:41:18] that is also very great. Um tell people about what you have built in this short period of time and uh we'll also be rewarding the top three to five people from our side who have whoever will get the most engagement on their app. Just

[3:41:31] make sure to tag hashtager and hash simply done in that post. Uh yeah I think um and if you want to share uh details regarding the program because I think most of the users were also having a little bit queries around there so we

[3:41:45] a little bit queries around there so we can take it up. you know firstly a huge round of applause for you. Um uh you know you

[3:41:59] walked us through um through uh you walked all of us through um you know a huge build process uh from like a blank screen to a working deployed AI agent

[3:42:11] live um in just a couple of hours and I think that's not a small thing to pull off um in a room this size. So thank you so genuinely um and uh yes I I'm seeing a lot of people who have built their agents here a lot of thumbs up and um um

[3:42:27] you so much. It's very incredible to see. Um and um you know that feeling right now watching something you built um you know actually respond and actually work. Um I think everyone each and every one of you here who got that

[3:42:42] that's exactly what this workshop was supposed to give you. Um however like you know our program is built to give you that same feeling 10 times over across you know the 10 weeks that the program spans for. Um and more than once

[3:42:57] a week you build something like this or work on some practical projects in addition to learning a lot of concepts. Um so you know if what you felt just now is something that you want more of then this is the program for you. um the

[3:43:11] professional certificate program in agentic AI and multi- aent systems and um you know as mentioned before it is brought to you in partnership with IIT partners uh wish wish list and high hub foundation um and the one guarantee we

[3:43:25] you will graduate from this program with workplace ready skills and we will look into exactly what these skills are um shortly I know su has taken through a lot of the details um so far um and I won't take like too much of the time.

[3:43:41] I'll just like briefly cover what the highlights of the program are and if you're interested we can like you know discuss further on the enrollment discuss further on the enrollment process. Okay. So um here's one thing

[3:43:53] that we that is very truly special about the program its duration. Firstly it's 10 weeks and that's it. And I say that because most programs um you know promising to teach you all of this um you know proper multi- aent systems rag

[3:44:06] MCP real production ready skills um usually ask for 6 months sometimes even a full year of your life. Uh 10 weeks is just just over like you know 2 months and that's a smaller time investment than most people put into like a single

[3:44:20] se semester of anything right and it's not a watered down 10 weeks either. It's live um instructor-led cohort-based learning. Um you're in it with people interacting with peers, mentors in real time through a dedicated Slack group. Um

[3:44:37] there are classes on the weekend, Saturdays and Sundays um and evenings. So if you work around um you know the whole schedule works around your job and everything else going on in your life and if you're um if even if you miss a

[3:44:49] class where there are recordings available um so that you never actually fall behind your cohort. Um and just because it's less time definitely doesn't mean like less value in that same window. You're hands-on with 25

[3:45:02] same window. You're hands-on with 25 plus tools um and um you know tools um that are actually used in the industry right now. um working you'll be working through 40 plus demos in real projects and um you know there will be a capstone

[3:45:15] project as well and you walk out with 15 plus skills that are genuinely in demand in the market right now. Um so really stick with that trade. Everything you would expect from a much longer program is compressed into just 10 weeks and

[3:45:27] it's backed by Microsoft as the industry partner and that's a serious return um you know for a small time commitment right and um here's the list of school uh skills and tools this that this program covers there are 15 plus skills

[3:45:41] and 25 plus tools all up on your screen um you know instead of going through each of these I just want to say like open any job portal right now and search for roles like agent KI engineer or AI product manager or genai engineer at a

[3:45:56] tech company in India today. Um and um you know you can try it now or try it later. I promise you at least five of these skills on the screen. Um something like you know multi- aent system or prompt engineering or rag or MCP or

[3:46:09] workflow automation they will be sitting right there in the requirement section and not like you know in a nice to have uh way but in the requirement. And here's the part um you know people don't expect this isn't just for engineers and

[3:46:23] tech people anymore even marketeers business analysts project managers even these job descriptions are starting to list the same skills right now right like agentic agentic AI has stopped being a tech team thing um it's the same

[3:46:37] story on the tool side as well lang chain and lang graph crew AI autogen chain and lang graph crew AI autogen lovable claude um and nin um all of these have become come um you know a very essential skill and a very

[3:46:52] essential tool to have in your resume right now. So this slide really isn't just a list of things that you will learn from the program. It's um you know closer to a checklist of what's already sitting in job descriptions today and by

[3:47:05] the end of this program you you would have actually built this um with um you single one of these tools so you can like you know um confidently add them in your resume and your profile. Um and yes with all the learning the skills the

[3:47:21] tools the projects aside here's something solid that you walk away with from the program the certificate. Um you might not realize how important this actually is. Uh so let me tell you you probably you know you would have seen a

[3:47:33] probably you know you would have seen a lot um of um online programs um and when you finish a course what you get is some co-branded certificate or sometimes not even properly properly from the university or the institution at all and

[3:47:45] that's where this program really stands apart. This certificate comes solely apart. This certificate comes solely from IIT partner and um it's IIT partner's name on it and that carries a very different kind of weight where a

[3:47:57] recruiter or a you know um hiring manager looks at it. It says you went through a program that um an institution like IIIT is willing to put its own name u behind fully and alongside that you also walk away with Microsoft uh um

[3:48:12] learn badges on the MSA learn portal. So you've got that industry recognized layer as well sitting right next to your IIT credential. So when this program ends, you're holding something that um actually helps you stand apart the

[3:48:25] moment someone uh sees your profile. And yes, here's a question we get uh all the time. Um honestly, one of the most frequent ones. How do we make sure that the curriculum stays relevant? Right? Because um you know, let's be real.

[3:48:39] There's a new AI tool. there is a new AI platform, a model, a new way of working that is coming out there every single week. So, uh what's very cutting edge today can feel outdated in just a month's time. So, that's exactly why we

[3:48:53] built AI impact studio. So, every 15 days we drop to um you know brand new live sessions there. Um they are not recordings or not something pre-back months ago. They are live expertled workshops just like this one. Um we

[3:49:07] built two clear tracks here. If you are um the builder type, the one who wants to get hands-on, ship things, automate workflows, um then you have an AI builder studio. Um if you're more the decision maker, the strategist, the one

[3:49:21] outcomes, then we have the AI business impact studio for you. Um so whichever always something fresh waiting for you. And the faculty teaching these sessions aren't just instructors orademicians. Um these are practitioners who have

[3:49:36] actually built and scaled AI inside companies like uh IBM, Amazon, Salesforce and a lot more. Um and here's the best part. If you are an active enrolled learner of this program, this this comes to you at no extra cost. It's

[3:49:49] this comes to you at no extra cost. It's already yours. And um um definitely the program doesn't stop at just the technical layer. One of the things that separates um uh Simply Learn from a from any other course platform is what

[3:50:02] happens after the learning. This program includes structure career support, interview prep and mock assessments tailored to AI roles, AI powered profile optimization for your resume and LinkedIn and handpicked job

[3:50:15] opportunities curated to match your skills. Um, as well as group mentoring sessions with industry experts. Um, because the credential that you earn from this program matters a lot. The portfolio matters a lot. But also what

[3:50:27] matters is the network that you build while you're earning um all these things, right? The people who move faster uh faster um than others in this AI field are um not always the most technically advanced. They are the ones

[3:50:42] who can articulate what they've built um who they've built it for and what it does, right? So this program is designed to make you that holistic uh person. And to make you that holistic uh person. And like Suash mentioned um you know uh we

[3:50:55] have an exclusive workshop uh discount available. Let me just take a moment to share the link. of the program. So you can go through all the details of the course here. Um the program is normally uh priced at 1

[3:51:12] lakh 19,9.99. You can go to the program page link, scroll down um and check the price for yourself. But because you showed up and built something great along with our experts, you get it at a 35% discount and that brings it down to

[3:51:26] just 78,000 rupees. Um the discount is live for 48 hours from right now. So if this is something you're even a little serious about uh tonight and tomorrow is the window for you. But we do understand that this might seem like a huge

[3:51:40] investment to make to some of you which is uh which is why we have a no interest EMI option starting at just about like 5,000 rupees a month. So I'm going to give you um four um you know quick recapping points to tell you why this is

[3:51:54] career and why it's totally worth it. One, the program is just 10 weeks. Two, the certificate is from IIT partners visualization uh IHub um and is not weight in the job market right now. Uh three, you get the AI impact studio new

[3:52:11] session every 15 days from people who have built um you know AI at uh companies like Amazon, IBM and so on. And four, it's hands-on learning throughout ending in a real um you know capstone project. And these projects and

[3:52:24] builds are going to make your resume extremely solid. So once again, do not miss this discount window and avail the 35% offer. The best investment you can make is on yourself and your knowledge. So we hope you take the right decision

[3:52:37] forward. Um you can also scan this QR code on the screen to know more about the program. You can download the syllabus, uh schedule a call with our our learners and then come to a decision

[3:52:50] to enroll. So please do check it out. With that, I'm going to launch a poll to take your interest in enrolling in the program. Please do note that this is not the poll for the certificate that is coming after this. This is a poll to

[3:53:02] take um you know your interest in enrolling in the program that we just spoke about. All you have to do is click on yes or no. Even if you have any queries, you need more details, you can um you know, you are um a little unsure

[3:53:17] and you want to talk to um someone to figure things out, you can click on yes. Our expert advisor will be getting in touch with you um you know, as early as tomorrow to guide you on the way forward. In case if you are answering

[3:53:30] in the chat why you're voting for no. We would love to hear um you know um more So please do feel free to let us know what's holding you back. what's holding you back. Okay. So, I'm going to have that uh poll

[3:53:45] live for a few more minutes right now. Um and uh uh so I think you took a lot of questions during the session itself. Um so if there's anything else that you

[3:53:57] would like to you know that uh um that you were not able to cover before and any final piece of advice that you would like to share with everybody um your closing thoughts that would be it would be great.

[3:54:11] Yeah, I mean uh whatever we tried out in this section that was very deterministic flow right because we wanted to explain you how the platform work and how the uh architect how you can use architect in

[3:54:24] studio to build out the use case but uh you will get a real feeling of the power solve certain kind of use case or problem that you're facing in your real life so I would say better to uh do two things first of all as Ana mentioned the

[3:54:39] basics of uh what the agents can do and how to build agents and everything. basics, you can actually come to a platform like architect or studio or any other platform which guys use and figure out a certain kind of problem, certain

[3:54:53] trying to do in your day-to-day life and try to solve it using these kind of platforms. then you'll get the kick around how you can use or build aentic applications and how you can uh make it a part of your life and you'll be able

[3:55:06] to better you know uh get the wow factor around the using these platforms and aentic because some of you might relate with the use case that I was teaching you about some of you might not but if you build something which you uh really

[3:55:18] think it would be a very crazy moment for you u I think yeah and uh last which I wanted to mention is don't forget to post whatever you have built built on LinkedIn and in Twitter tag # Lizer and tag # simply learn so that we would also

[3:55:34] get to know what you guys have built out. Thanks. out. Thanks. >> Yes. Yes. So, thank you so much. Um and uh I just want to address a couple of questions. Uh there's someone asking if

[3:55:46] you need a coding background for this program. Absolutely. Um you know, not required. There is a Python refresher module uh module um that is optionally available in the program. So if you don't have um you know any coding or

[3:55:59] programming knowledge then you can take up that module to um you know sort of uh um understand the foundations and the basics. Um otherwise it is not mandatory Um there were a couple of people who had asked about the USD pricing. Um please

[3:56:15] do like you know um just answer yes in the poll. uh you will be contacted by our expert advisor who will let you know the details um the specific um you know uh price of the course according to um the country that you are from we will be

[3:56:30] able to help you out with all the details um there um there's someone here who's already enrolled in the course um and um um he has shared I think subroy

[3:56:42] um he shared that the course is very good and the best part uh is that it is good and the best part uh is that it is 100% live So great um amazing to see all of the interest that has come in. Um I think more than uh uh 200 of you have uh

[3:56:57] um answered the poll which is great. So I'm going to end the poll right now and uh yes like uh Syash had mentioned we would love to see um you know what you have uh built so far. This is um honestly one of my favorite bits of the

[3:57:12] whole workshop as well. Um so if you didn't just like you know sit and watch and you actually built and deployed a working AI agent um it's genuinely a big leave here without um you know an opportunity to show that off. Um so

[3:57:26] first you can open a folder um open the folder link in the chat. Let me just folder link in the chat. Let me just share um that.

[3:57:44] or you can scan the QR code. It will take you to a Google Drive folder um where you can just upload a screenshot of the agent or like video um whatever like you know you would like however you would like to present the agent that

[3:57:56] you've built. You can just um you know record that and um upload that in um the folder. Um just make sure that we have your uh name and email id correctly so out to you we would be able to uh do that. the best uh three subitions will

[3:58:12] get featured on Simply Learns in Instagram where we have 200,000 plus followers. Um so you know if if you built something uh great with Sri today, it. Make sure that you everybody everyone knows about it and make sure

[3:58:28] that you get rewarded for it because as Sash mentioned um if you tag uh Liza and Sri Learn the post with the highest engagement uh actually in fact a few a couple of posts at least with the highest engagement wins uh Liza's uh

[3:58:41] swag kit um and some goodies from us as well. So um yes. So let u um you know I'm just going to keep this slide on for a couple of uh more minutes so that you

[3:58:54] a couple of uh more minutes so that you can um you know scan the QR code. Yes, the folder is empty right now. You would have to upload uh you know whatever you built James in that folder. just ensure that the name of the file is

[3:59:09] current um you know has your email id um so that we will be able to reach out to so that we will be able to reach out to you right okay great with that um I will move on to the next and the final part of the workshop we know many of you

[3:59:25] attending certificate I am launching another poll right now where you can enter your full name um in this poll you will have to um Just give us your full

[3:59:37] name. Do not share your email ids or do not add any title to it like Mr. Miss not add any title to it like Mr. Miss etc. Um all you have to do is just um you know type your full name um the name that you want on the certificate and

[3:59:50] click on submit. We will be able to generate the certificate and share with you on email within the next 48 working hours. And along with the certificate, you will also be receiving the deck that we presented throughout um you know this

[4:00:03] session um and the GitHub link that was shared um in the session as well as the recording of the session. So you will be getting all this as um you know you attended the workshop and you were um you know part of the session with us.

[4:00:19] Okay. So I am going to have that live for uh just a minute more and uh yes so with that we are we have come to the end of the session. Once again thank you so much to every single one of you for being here for the last 4 hours um you

[4:00:35] know building along with us. Um it I think it was a real commitment and it says a lot about what you're going to build forward. So we hope that you would be able to apply the learnings from this session um as well as what you will get

[4:00:48] from the program uh to get ahead in your career. We hope to see you see a lot of career. We hope to see you see a lot of you enrolled um and starting the next uh um you know cohort of the program very soon. So have a great day everyone. Once

[4:01:02] again uh thank you so much uh to Suryash and um Anerud and the whole team of Liza for um you know helping us with this workshop. It was amazing and we hope to

[4:01:14] you know um do another session once again uh very soon. All right. So if anyone has any concerns or any questions that we were not able to address, please do um write to us at

[4:01:27] [email protected] In case if you're not able to access the poll or um you know your certificate does not arrive in your inbox uh by this weekend just write to us at webinars at simplylearn.net and we will be able to

[4:01:40] look into it and get back to you. So with that we're wrapping up the session. Thank you so much everyone. >> Thank you.