---
title: 'What Is ChatGPT 5.6? | GPT-5.6 Explained | ChatGPT 5.6 Tutorial For Beginners | Simplilearn'
source: 'https://youtube.com/watch?v=uWueU6hcgLE'
video_id: 'uWueU6hcgLE'
date: 2026-08-08
duration_sec: 1990
---

# What Is ChatGPT 5.6? | GPT-5.6 Explained | ChatGPT 5.6 Tutorial For Beginners | Simplilearn

> Source: [What Is ChatGPT 5.6? | GPT-5.6 Explained | ChatGPT 5.6 Tutorial For Beginners | Simplilearn](https://youtube.com/watch?v=uWueU6hcgLE)

## Summary

This video from Simplilearn provides a comprehensive overview of OpenAI's GPT-5.6, released on July 9, 2026, as a family of three models: Soul, Terra, and Luna. It explains how GPT-5.6 moves beyond simple question-answering to support complete workflows across different environments like ChatGPT, ChatGPT Work, and Codex. The tutorial includes practical demonstrations of model comparison, business analysis, and dashboard creation, emphasizing the importance of verification and human oversight.

### Key Points

- **GPT-5.6 Overview** [00:07] — GPT-5.6 is no longer just a Q&A tool; it can handle larger assignments, work across files and applications, and turn goals into finished materials like spreadsheets, presentations, reports, or web apps. Released July 9, 2026, as a family of three models: Soul (flagship), Terra (balanced), and Luna (fastest and most cost-efficient).
- **Workflow Capabilities** [01:01] — GPT-5.6 can complete connected sequences of work, e.g., reading campaign data, checking meetings, calculating results, creating a management presentation, and building a dashboard from the same information. This matters because people use these tools for far more than casual questions.
- **Ecosystem Fit** [03:42] — GPT-5.6 represents the underlying intelligence, while different environments serve different tasks: ChatGPT for quick answers, ChatGPT Work for multi-step workflows with files, and Codex for software development tasks. The key difference: ChatGPT answers questions, Work completes larger workflows, Codex builds and modifies software.
- **Three Models Explained** [05:44] — Soul is the highest capability model for deep reasoning, advanced problem-solving, research, and complex coding. Terra is the balanced option for everyday professional work like reports and presentations. Luna focuses on speed and cost efficiency for simpler, repetitive tasks. The best model depends on task, quality, and resources.
- **From Prompts to Workflows** [07:39] — Modern AI usage shifts from simple prompts to complete workflows. Instead of 'analyze this marketing data', a better request is 'review the campaign data, identify best-performing channels, compare results, and prepare recommendations'. Defining the goal and output clearly helps AI produce more useful outcomes.
- **Verification and Limitations** [10:54] — A common mistake is assuming confident-looking answers are accurate. AI can create impressive results, but important decisions require human review. AI can only work with the information it receives; if data lacks profit margins, AI cannot determine profitability. Responsible usage includes reviewing shared information, especially confidential data.
- **Input Files Preparation** [13:05] — The tutorial uses several input files: campaign_performance.csv, marketing_meeting_notes.txt, executive_brief.txt, business_objective.txt, and presentation_reference_input.txt. These files contain campaign data, meeting notes with a deliberately incorrect claim about webinars, and other context for the workflow.
- **Choosing the Right Environment** [15:22] — Three requests illustrate environment selection: explaining a concept (regular ChatGPT), creating a spreadsheet/report/presentation from files (ChatGPT Work), and building a dashboard (Codex). The choice depends on whether the task is a simple question, a multi-step workflow, or a software development project.
- **Model Comparison Demo** [22:12] — The demo compares Soul, Terra, and Luna using the same two files and prompt. Evaluation categories include calculation accuracy, insight quality, recommendation quality, presentation quality, corrections required, and practical value. The goal is not to prove one model always wins but to find which reaches required quality with acceptable time, cost, and correction.
- **ChatGPT Work Demo** [27:14] — A new work project is created with the four files. The system is asked to inspect materials and create a plan before execution. The plan includes checking meeting notes against the CSV, identifying the incorrect webinar claim, and creating a table with statuses: verified, incorrect, or unsupported.
- **Spreadsheet and Presentation Generation** [31:21] — Using the same resource material, the system generates a spreadsheet with calculated metrics (cost per lead, cost per enrollment, lead-to-enrollment conversion, return on ad spend) and an executive presentation. The final sheet preserves original data, and the second sheet contains calculated metrics.

### Conclusion

GPT-5.6 represents a shift from simple AI assistants to workflow-oriented tools that can handle complex, multi-step tasks across different environments. The key takeaway is that while AI provides speed and assistance, human judgment, verification, and responsible usage remain essential for reliable and accurate outcomes.

## Transcript

line and it's no longer being positioned as only one place where we tap a question and receive an answer. With GPD 5.6 and Chad GPD work, it can now stay with a larger assignment, work across files and applications, and turn a goal
into finished material such as a spreadsheet, presentation, report, or a web app. OpenAI released GPT 5.6 6 on July 9th, 2026 as a family of three models. Soul, the flagship model for the hardest work. Terra, a balanced option
for everyday professional tasks. Luna, the fastest and the most costefficient option. So when we say charge GPD 5.6, with a different number. We are talking about a new set of models powering
about a new set of models powering different ways of working inside chargi that the tool can move beyond just writing replies and help a complete
connected sequence of work. For example, reading campaign data, checking meeting calculating results, creating a management presentation, and then using the same information to build a dashboard. Now, this matters because
people are already using these tools for far more than just casual questions. So, now that we can understand why chart GPT 5.6 matters, let's move on to the agenda. GPT 5.6 6 overview understand what GPT 5.6 6 is and why it's so
important. Soul, Terra, and Luna. Compare the three model options and their use cases. Chart GPD work and codeex. Learn which environment suits each type of task. Key improvements. Explore what has changed in reasoning,
coding, design, and workplace tasks. A comparison demo. Test sold Terra and Luna using the same campaign data. Charg work demo. Turn raw files into analysis spreadsheet and presentation. Codeex dashboard demo. Build a working campaign
verification and limitations. Check claims, formulas, privacy risks, and on, let me share something really important with you. If you're actively looking for jobs or planning a career switch or worried that your current
skills may not be enough for the next wave of AIdriven roles, then learning how to actually build and apply AI can give you a much stronger advantage than So the professional certificate course in generative AI and machine learning
from EICA consortium IIT Kpur is in collaboration with simply learn and is also helping you to design job relevant skills across Python, data science, machine learning, deep learning, generative AI, prompt engineering, NLP,
reinforcement learning and agentic AI. You get hands-on exposure to more than 30 tools and libraries including Python, TensorFlow, PyTorch, Langchain, GitHub Copilot, Open CV and charg. So you're not just learning concepts. This program
also includes more than 18 real world projects across marketing, finance, healthcare, HR recommendation systems, computer vision, and generative AI, helping you create a portfolio that can support job applications, interviews,
and career transitions with live expertled training, IIT Kpur Masterclass, hands-on workshops, and career support. This program is working professionals, developers, analysts, and career switchers who want
a structured path into AI and machine learning roles. Now, before we move ahead, here is a quick question to test your understanding. Which GPD 5.6 model is designed as a balanced option for everyday professional work? Is it a
Soul, B Terra, C Luna, or is it D Codeex? Let us know your answers in the comments below. So, before we explore the capabilities of GPT 5.6, six. Let's first understand where it fits in the AI ecosystem. So when people hear GPT 5.6,
chatbot version. But there is something more to it. So GPT 5.6 represents the underlying intelligence while different environments like charg different types of tasks. So a normal charge GPT conversation is useful when
you need luck, answers, explanations, brainstorming or content assistance. But professional work usually involves multiple steps, files and decisions. And this is why chart GPT work becomes so useful in helping users analyze
information and creating outputs like reports. And codeex focuses on software development tasks where the requirement is not just explaining code but actually working with projects, creating files and building applications. So the key
difference is simple. Chad GPT helps you answer questions. Charge GPT work helps you complete larger workflows and codecs can help you build and modify software. So now that we understand where GPT 5.6 fits, let's look at what makes this
generation different from traditional AI assistance. Now the best change is the ability to handle more complex tasks and involve multiple steps, different sources of information and practical outcomes. So earlier many used AI mainly
for individual tasks like writing an email, summarizing a document or answering a question. But real work is rarely limited to only one action. So business analysts may need to review reports, find insights, create
recommendations and present the findings to leaders. GPT focuses on supporting these larger workflows. It can help with analyzing information, work with files, create professional documents, assist with coding tasks, and support
decision-m process. However, the goal is not just to replace human thinking. The real value comes from combining AI's ability to process information quickly with human judgment and experience. So, for example, AI can help identify
marketing professional still needs to for the business. So, this shift from simply generating answers to helping complete meaningful tasks is what makes JA GPT 5.6 so important. So, now that we
understand the overall improvements, let's have a look at three GPT 5.6 6 models and understand when each one makes sense. GPD 5.6 introduces three different models. Soul, Terra, and Luna. Each model is designed for different
levels of complexity, speed, and efficiency. So, Soul is the highest capability model built for tasks that require deeper reasoning, advanced problem solving, research, and complex coding. For example, analyzing multiple
complicated technical challenge would benefit from a model like Soul. Now, Terra is the balanced option designed for everyday professional work. Tasks like creating reports, analyzing documents, preparing presentations, and
solving common workplace problems fit well here. Now, Luna focuses on speed and cost efficiency. This is useful for simpler repetitive tasks where organizations use to hand large number of requests. Now, the important point is
that the best model is not always the most powerful one. The right choice depends on the task, the expected quality, and the resources available. Selecting an AI model is similar to choosing the right person for a job. Not
expertise. For example, if a company needs to classify thousands of customer messages, a fast and efficient model may complete the task effectively. But if the company needs to analyze complex
financial documents and prepare strategic recommendation, a more advanced model may be a better choice. When comparing models, we should always look beyond the final answer. Its important factors include accuracy,
speed, cost, and how much human correction is required after the output is generated. So a model that produces a longer response is not automatically better. The real question is whether it provides the right result for the
specific requirement. And this approach helps businesses use AI more efficiently instead of choosing one model for every situation. So now that we understand how to select models, let's move on to see a practical example and see how GPT 5.6
can handle a business workflow. So one of the biggest changes in modern AI usage is moving from simple prompts to complete workflows. So basic prompt may ask summarize this report. While useful, it solves only one small part of a
larger task. In a real business environment, the requirement is usually much broader. A team may need to analyze multiple documents, compare information, identify problems, create reports, prepare presentations, and make
recommendations. And this is where the way we communicate with AI becomes so important. Instead of only asking for an answer, we need to be explaining the goal, provide the necessary information, and define what the final output should
be achieving. Let's say for example, instead of saying analyze this marketing data, a better request would be review the campaign data, identify the best performing channels, compare results, and prepare recommendation for the next
campaign. Now, this gives AI more context and helps produce a more useful outcome. So, the better we define the problem, the better AI can support the workflow. Now let's have a look at how chart GPT work can help convert raw
information into useful business outputs. So imagine a marketing team has a spreadsheet containing campaign performance, meeting notes, discussing explaining what leadership wants to know. The challenge is not only
The challenge here is to connect all of the information together and create something useful. Chart GPT work can help analyze these inputs and transform them into outputs such as reports, spreadsheets, presentations, and
data can reveal which channels are generating the highest revenue, which channels were most efficient, and where improvements may be required. But a good analysis also requires understanding limitations. So, if available data does
not include customer lifetime values, we cannot be making claims about long-term customer profitability. And if there is no audience information, we cannot assume which platform generated better quality leads. And this is why combining
AI capabilities with human review is so important. AI can organize and analyze information, but humans still provide the business context and final decision-m. So now that we have understood how GPT 5.6 can help analyze
information and create professional outputs, let's explore an other applications with codecs. So while traditional AI assistants were mainly used for answering questions or generating content, Codeex focuses on
workflows. So this helps developers understand existing code, create application files, modify them and accelerate the development process. Say for example, imagine a marketing team has analyzed campaign data and wants to
interactive dashboard where users can view spending, leads, revenue, and campaign performance. So instead of manually creating every component from scratch, developers can describe the requirement and codeex can assist in
creating the structure, suggesting improvements and supporting different development tasks. However, Codex is not about simply generating code and Developers still need to understand the requirement, verify the generated code,
test the application, and improve the final solution. The real value comes with combining AI assistance with developer expertise, allowing teams to reduce repetitive work, and spending more time focusing on solving complex
problems and building better applications. So, while GPT 5.6 can generate reports, analyze information, and create useful outputs, one of the most important steps in any AI workflow is verification. A common mistake people
make is assuming that a confident-look answer is always accurate. AI can create impressive results, but important decisions still require a human review. For example, imagine an AI system analyzing campaign performance and
suggesting that webinars generated the highest revenue. Now, before making a comparing this conclusion with the original campaign data and verify whether the numbers actually support that claim. Now, the same applies to
different AI generated outputs. A spreadsheet may contain an incorrect calculation. A presentation may include an unsupported statement or a dashboard may display the wrong metric. Now, a reliable AI workflow does not end when
the output is generated. It includes checking information, validating important conclusions, and improving the final result before it's used. The best outcome happens when AI provides speed and assistance while humans provide
judgment, experience, and decision-m. Understanding its limitations is equally important. AI can only work with the information it receives. So if important provide the complete picture. For example, if a campaign data set contains
spending and revenue information but does not include profit margins, AI can identify revenue performance but it cannot confidently determine which campaign generated the highest profit. Another important factor is human
judgment. AI can identify patterns and suggest recommendations. But business decisions often require understanding company goals, customer behavior, and practical limitations. So responsible usage is also essential when working
carefully review what information they share, especially while handling confidential business data, customer information, or sensitive documents. trust AI or to avoid using it completely. The most effective approach
is to use AI as an assistant that can help people work faster and make better decisions while maintaining accuracy, privacy, and accountability. Now before prepare the input files that we will be using throughout the workflow. So you
can head over to any ID of your choice and I've chosen VS code for my convenience. So here you go and then you you move on to file and then create a new text file and save it as campaign performance.csv which is the first file
workflow. So now my CSV file will be containing the following input. So this that will be used for the model comparison, spreadsheet, presentation, let's go ahead and create the second input. So you create a new text file.
input. So you create a new text file. Give it the title of marketing meeting notes.xt. So this is a text file. So the input we follows. Google search brings the largest volume of enrollments but
requires the highest campaign spend. LinkedIn leads are believed to include more managers, but the current data set doesn't contain any job ro information. YouTube creates broad awareness, but the marketing team is uncertain about its
complete revenue attribution. Email has its strongest return, but the available mailing list is limited. Webinars appear to convert interested leads effectively, enrollments before increasing or without increasing the total campaign budget.
The marketing team believes that the webinars generated the highest revenue. deliberately incorrect. The campaign CSV shows that Google search generated the highest revenue and not webinars. So this incorrected statement is what will
help us test whether Chad GPT 5.6 checks the meeting notes against the original data set. Now let's add an executive brief input. We'll create a file for that. So this will also be saved as a text file and the input for this is as
follows. You can pause here to have a look for your reference. Now let's move on to business objective. Now let's go ahead and create a business objective. word, please? Now let's add a presentation reference input. So let's
go ahead and save this file as presentation reference input. And we will be saving this as a text file as well. The input for this is as follows. You can pause to have a look for your reference. So a
file with an existing presentation template or a brand reference. So now let's go ahead and understand what we are going to be learning. So before consider three different requests. The first request is explain the difference
generation. Now this is a normal question. It doesn't require several files, a complete application workflow or a finished business deliverable. So a regular charge GPT conversation is sufficient. Before we begin with the
different requests. So the first request is explain the difference between brand awareness and lead generation. Now this is a normal question and it doesn't require several files, a complete application workflow or a finished
business deliverable. So a regular charge GPT conversation is sufficient. campaign files and create a spreadsheet, a management report and an executive presentation. Now this is no longer a simple question. It's a complete
calculations, decisions and multiple outputs. So a work focused environment is more suitable. Now the third request is open the software project, inspect the existing files, build a dashboard and test whether it works. Now this
code files. So codeex is a relevant demonstration we should be able to decide whether a task belongs in a decide whether a task belongs in a normal charg.
between soul terra and luna and write a complete project prompt and verify the final output without using it. So I'm opening charge GPT and I've signed in. So at the top of a normal conversation I can open the model selection area. So as
our models from here. So the exact options visible may vary from depending on account subscription region and workspace settings. So a standard chat GPD conversation is useful when I want to ask a question, explain a concept,
brainstorm ideas, rewrite content or solve a problem conversationally. A work focused environment is useful when I have several files, a clear business objective or multiple deliverables. So instead of asking only one question, I
can describe the final result that I need. Now the system can review the supplied information, organize the work and help create connected outputs such as connected outputs and reports, spreadsheets and even presentations.
Now, Codeex is designed for software development tasks and I can inspect a project folder, understand existing code, create or modify files, help build can understand these environments through one marketing example where I
through one marketing example where I ask what is return on spending with advertising. So now I'm asking for an explanation. So a normal RGPT conversation is more than enough. Now when I ask use these campaign files to
when I ask use these campaign files to create a complete business review I am assigning a larger work project. Repeat I am assigning a larger work project. I am assigning a larger work project. Now when I ask turn this campaign data
Now when I ask turn this campaign data into a working website dashboard. So development task and that is where codeex becomes relevant. Now let's have a look at three GPT 5.6 model levels. Soul is the highest. Now this is
suitable for difficult assignments where deeper reasoning, stronger output more than selecting the lightest model. Terra is a balanced option for normal professional work and Luna is a faster and more efficient option for simpler
work, repeated transformation, extraction, classification and high remember this is to imagine a workplace team. Soul is like a senior specialist assigned to the most difficult problems. Terra is like an experienced generalist
handling everyday professional work and Luna is like a fast operations assistant processing simpler repetitive tasks. Now this is only a memory aid and the correct model should still be selected by testing it on an actual task. So
instead of reading a long list of benchmark numbers, let's understand the improvement through a business example. So from what you know I have a folder and a business brief. Now in a basic question and answer workflow, I might
upload only the campaign CSV and ask RGBT to summarize it. So let's first demonstrate that simple workflow. So as you can see I'm opening a normal RGBT conversation and attaching the campaign performance file. So go ahead and add
performance file. So go ahead and add photos or files. You move to documents. You're recording at 11 shots. your topic. [clears throat] Oh yeah, give me three minutes. I mean I I need I still have to record but I'll
give you in 3 minutes. So I will be entering the following prompt. Review entering the following prompt. Review the attached campaign data and brief the attached campaign data and brief which marketing channels performed well
which marketing channels performed well and repeat and which channels need and repeat and which channels need improve followed by use only the improve followed by use only the information available in the uploaded
information available in the uploaded three file. Do not make any assumptions three file. Do not make any assumptions about the audience quality, customer about the audience quality, customer lifetime, value or profit because those
lifetime, value or profit because those details are not included within this file. So this prompt contains three simple parts. The first part tells charge GPT to review the campaign. The second part asks for a strong and weak
prevents the response from inventing data set. After submitting the prompt, chat GPT should provide a useful summary. So this may not it may notice
that Google search produced the largest number of enrollments. Email was most costs are made for more leads or enrollments. So as you can see the results I just spoke about are as displayed over here. However, this still
may not be only a written answer. It's not yet a complete business deliverable and I would still need to create the formulas, check the tools, build the presentation and write the management summary. So the larger GPT workflow, so
the larger GPT 5.6 workflow is useful because I can describe the final outcome rather than just asking one isolated question. I can request a spreadsheet with formulas, charts, and executive presentation and a management summary. I
can also provide meeting notes and ask the system to check whether they agree with the source data and I can define repeat and I can define what must remain unchanged and request an audit before using the final result. Now the
practical improvement is still therefore not simply that charge GBD produces a longer response. The improvement is that we can organize our larger assignment around a final outcome while still providing reliable evidence constraints
and a completion standard. Now let's do a comparison between Soul versus Terra versus Luna. So for this comparison I'm creating three separate work conversations. In the first conversation I select soul. So there are two modes in
the so there are two modes in charg. First is the normal chat and the second one is work. So you will have to select on work. Now moving on in the first conversation I select soul. In the second I select error and the third I
select Luna. So let me go ahead and show you how to do that. But before doing but before moving on I have to let you know that the exact models visible to any account may vary but the main purpose of this demo is to show how a fair
comparison should be structured. So now I will go ahead and attach the same two files for all three conversations which is campaign performance.csv as well as the executive brief. So I use exactly the same prompt in every conversation
and keep the reasoning or quality setting as similar as possible. So now let me go ahead and add the following prompt to this chat. Analyze the above documents which I have attached a concise three
slide executive review. For slide one show total campaign spend, total leads, total enrollments, total revenue and overall
enrollments, total revenue and overall return on advertising spend. Slide two. return on advertising spend. Slide two. compare all of the five channels using.
So we have options here which is cost per lead, cost per enrollment, lead to per lead, cost per enrollment, lead to enroll version rate, return on
advertising spend. This takes a lot of time to type man. And I will be pasting the following prompt here. So we have the prompt for slide three and we have some important constraints listed over here followed by completion standard. So
this prompt defines the final outcome the content required on each slide the constraints and the completion standard. Now after three outputs are generated I don't automatically decide that one model is better close the door now
please so this prompt defines the final outcome the content required on each slide and the constraints and completion standard. generated, I don't decide that one model is better simply because the slides look
more polished. I compare all of the three using the same scorecard. So here are the evaluation categories. Firstly, we have the calculation accuracy. Are all of the totals correct? Are all the formulas correct? And does the total
proposed budget equal to 4 lak 80,000 rupees? Then we have insight quality. Does the output explain why a channel performed well? Does it separate scale from efficiency? Does it identify missing information and does it separate
quality. Is each recommendation supported by data? Does the recommendation consider practical limits? Does it avoid treating one high scalable? Then we have presentation quality. Is each slide easy to
understand? Are the most important numbers clearly visible? Do the slides follow a logical sequence? And do the slides follow a logical sequence? for corrections required, how many numerical corrections are needed, how many
supported, how many unsupported claims need to be removed and how many formatting improvements are needed. Finally, we come to practical value. Could a business leader understand the recommendation quickly? Is the output
ready to use or does it require minor revision or does it require a major revision? Now, let's go ahead and verify the expected totals manually. Now, total spend is 4 lak 80,000 rupees. Total leads as you can see are 7,250.
Total enrollments are 814 and total revenue is 24 lakh 64,000. So overall turn on advertising spend is total revenue divided by the total spend which is approximately 5.13. Now I compare the outputs. As you can see we have the
slides made over here. A basic response may simply mean that an email should be performing well. A stronger response should explain that an email generated the highest return on advertising spend but its performance may depend on the
size of the existing mailing list. So as you all know a basic response may say that the Google search is expensive but a stronger response should explain that number of enrollments. A stronger response should be explaining that
Google search also produced the largest number of enrollments and the highest revenue. Now a basic response may recommend the shifting of the entire budget to an email. A stronger response should be recognizing that the email
acquisition channels. So the purpose of this comparison is not to prove that soul will win every small task. A simple extraction task may be completed well by Luna. A standard professional analysis
may be handled well by terror and a complex assignment involving conflicting files, several deliverables and executive and executive quality output may benefit more from soul. So the correct question is not which model is
always the best. The correct question is which model reaches the required quality for this task with an acceptable level of time cost and correction. Now I'm creating a new work project with the name of GPD 5.6 campaign business. So
button over here. You can you can head over to the choose project button. Click on new project and then type your project name. So I will be typing the project name. So I will be typing the repeat GP campaign business review. So
now that the project is created, let's go ahead and upload the following files. So upload the following files. Campaign performance.csv, marketing meeting notes, execute a brief presentation reference. So now before requesting for
the spreadsheet, presentation or report, I first ask the system to inspect the material and create a plan. So first let's add our files. So once the files have been added, I go ahead and type the following work planning prompt. So
following work planning prompt. So review all of the four documents review all of the four documents attached above before creating any final attached above before creating any final followed by first give me a short work
plan showing let me go ahead and paste the following prompt for your reference. So I've asked it to repeat. So I've asked it to go ahead and review all of following. Repeat with the following points to keep in mind. Now this prompt
asks for a plan before execution. Identifies the campaign CSV as the main numerical source and prevents the system from creating several deliverables before we review its approach. Now the work plan should be including total
spend, total leads, total enrollments, total revenue, cost per lead, cost per conversation rate, lead to enrollment conversion rate, return on advertising spend, share of total enrollments, and share of total revenue. It also mentions
that the claims in the meeting notes will be checked against the CSV. Now, the system should be identifying the statement claiming that the webinars generated the highest revenue. The data set shows that this is incorrect because
Google search generated 8 lakh 64,000 rupees in revenue while webinars generated 3 lak 64,000 rupees. Now if the contradiction is missing, I will taken this prompt from GPT for your easy reference. So this is what it says.
numerical and performance related statement in the meeting notes with the campaign CSV. Create a table with these columns. Meeting note statement supporting CSV value status whether it's verified incorrect or unsupported
explanation explanation repeat explanation corrected wording when required use the CSV file as a source of truth for numerical performance do not create the final deliverables yet so this branch separates three different
situations verified means that the source supports the statement incorrect statement and unsupported here means the files don't contain enough information now the link England statement should be marked unsupported because data set
because the data set doesn't just contain job ro information. The webinar incorrect. So once the plan and contradiction check are correct I approve the analysis. So now let's go ahead and carry out the complete
campaign analysis. So the prompt for the same is as follows. I got this prompt from charge GPT. So as you can see here are the points which have been included where we have we supposed to calculate the following for every channel along
with that we have identify and recommendation rules. So the expected as the channel with the greatest enrollment scale because it produced repeat yeah because it produced 288 enrollments. Google search also
generated the highest revenue at 8 lakh 64,000 rupees and email repeat and email has the strongest return on advertising spend because 5 lakh 40,000 rupees divided by 20,000 rupees equals 27. Now webinars have the highest lead to
enrollment conversion rate because 91 divided by 650 equals to 14%. Now YouTube generated a large number of leads because its enrollment and revenue efficiency are weaker than email, Google search and webinars. LinkedIn may still
reach a useful professional audience but the files don't prove that the leads have a high seniority or a long-term value. Now this is important because a good analysis should not be turning a team belief into a measured fact. So we
demonstration part here in charge GPT with this project. Now let's go ahead and use the same resource material with the spreadsheet executive presentation business deliverables. So I use the following prompt. So I got this prompt
reference. So this has the following which is spreadsheet requirements, calculated metrics, included charts, some points for being able to clearly distinguish and the other instructions are as follows here. So this prompt
deliverables and defines what must be checked before the work is complete. Now with the spreadsheet. Now the final sheet should preserve the original campaign data and the second sheet should be containing calculated metrics.
Cost per lead is spend divided by leads. Cost per enrollment is spent divided by enrollments. Lead to enrollment conversion is the enrollments divided by leads. Now return on advertising spend is revenue divided by spend. Now for
Google search cost per lead is 1 lak 80,000 rupees divided by 2400 which equals rs75 and cost per enrollment is 1 lak 80,000 divided by 288 which equals 625 rupees and for email cost per lead
is approximately 11 rupees and cost per enrollment is approximately 92.59 rupees for webinars lead to enrollment conversion is 14%. And that brings us to
the end of this tutorial on GPT 5.6. We explored soul, terra, and luna and understood the difference between chat GPT work and codeex and saw how to turn raw information into useful business outputs while checking every important
more such insights and check out our related videos.