[0:00] I don't usually overhype myself but in
[0:02] this video we're going over the only
[0:04] formula you will ever need to master
[0:06] prompting on ChatGPT and Google Bard
[0:08] so let's get started hey friends welcome
[0:11] back to channel if you're new here my
[0:12] name is Jeff I work full-time in Tech
[0:14] and if you're anything like me a couple
[0:15] months ago you know prompting is an
[0:18] important skill to learn but you're not
[0:19] exactly sure why some prompts generate
[0:22] outputs that are super generic While
[0:24] others give you precisely what you're
[0:26] looking for since then I've spent
[0:27] hundreds of hours taking prompt
[0:29] engineering courses and applying what
[0:31] I've learned in my daily life and so in
[0:33] this video I'm sharing the six building
[0:34] blocks that make up a good prompt so
[0:36] that you can use this formula to
[0:38] consistently generate high quality
[0:41] outputs first it's critical to not only
[0:43] know what the six components are task
[0:45] context exemplars Persona format and
[0:47] tone but also know that there's an order
[0:50] of importance to these six components to
[0:53] show you what I mean let's use this
[0:54] simple example I'm a 70kg male give me a
[0:57] three-month training program the first
[0:59] part is context followed by the task the
[1:01] reason why the task is higher up in the
[1:03] form of the hierarchy is if we just
[1:05] input the task without the context
[1:06] there's still some sort of meaningful
[1:08] output but if we just give ChatGPT the
[1:11] context nothing really happens put
[1:13] another way it's mandatory to have a
[1:15] task in your prompt it's important to
[1:18] include relevant context and exemplars
[1:20] and it's nice to have Persona format and
[1:23] tone when you think of writing your
[1:25] prompt go down this mental checklist so
[1:28] this formula will act as a constant
[1:29] reminder for you to include just enough
[1:31] relevant information when writing
[1:33] prompts and as you'll see in this next part
[1:35] you do not need all six components in
[1:38] every prompt to have a good output now
[1:40] let's break down each building block
[1:42] with specific examples starting with the
[1:44] task the rule of thumb is to always
[1:46] start the task sentence with an action
[1:48] verb generate give write analyze Etc and
[1:51] clearly articulate what your end goal is
[1:54] it could be one simple task like
[1:56] generating a three-month training
[1:57] program or a complex three-step ask like
[2:00] analyzing hundreds of user feedback
[2:02] sharing the top three takeaways and
[2:05] categorizing the feedback based on the
[2:07] team responsible for following up the
[2:09] second component context is the
[2:10] trickiest to get right because
[2:11] technically there's an infinite amount
[2:13] of information you can give so I found
[2:16] asking myself these three questions to
[2:18] be super helpful in coming up with just
[2:20] enough information to get a good result
[2:22] from ChatGPT first what's the user's
[2:24] background second what does success look
[2:27] like and third what environment are they
[2:29] in Back to the workout example we now
[2:31] have I'm a 70kg male looking to put on
[2:34] five kilograms of muscle mass over the
[2:35] next three months I only have time to go
[2:37] to the gym twice a week and for one hour
[2:39] each session give me a three-month
[2:41] training program to follow could I have
[2:43] added more background information of
[2:45] course only prioritize the muscle groups
[2:47] that make me look good on Instagram
[2:49] but the key to staying productive with
[2:51] ChatGPT and Bard is giving just enough
[2:53] information to constrain the endless
[2:55] possibilities by the way although this
[2:57] video is not sponsored it is supported
[2:59] by those of you who subscribe to my paid
[3:01] productivity newsletter on Google
[3:02] workspace tips Link in the description
[3:04] to learn more moving over to the
[3:06] exemplars component it's just a fancy
[3:08] way of saying examples basically all the
[3:10] research on large language models LLMs
[3:13] have shown that including examples
[3:15] within the prompt drastically improves
[3:18] the quality of the output starting with
[3:20] a simple example this is a poorly written
[3:22] bullet point from a resume we can now
[3:25] ask ChatGPT to rewrite this bullet
[3:27] point using this structure I accomplished
[3:29] X by the measure y that resulted in Z
[3:31] which is actually best practice by the
[3:33] way so actually do this in your resume
[3:34] for example I lowered Hospital mortality
[3:37] rate by 10% by educating nurses in new
[3:39] protocols which translates to 200 lives
[3:41] saved per year here's a slightly more
[3:43] complicated example for interview prep
[3:45] based on my own resume write me an
[3:47] answer to the interview question what's
[3:49] your biggest weakness use the star
[3:51] answer framework situation task action
[3:53] and results here instead of using a
[3:56] full-blown interview answer as an
[3:57] example which would be overkill the star
[4:00] framework acts as an example structure
[4:02] for ChatGPT to follow last example
[4:04] let's say you need to write a job
[4:05] description you give some context around
[4:07] the opening and ask ChatGPT to
[4:09] reference an existing job description if
[4:12] I use this one I found on LinkedIn the
[4:13] output will follow the same formatting
[4:15] and use the same professional HR-y
[4:17] language saving me a bunch of time main
[4:20] takeaway here exemplars are not
[4:22] necessary for every prompt but including
[4:25] a relevant example or framework will
[4:27] greatly improve the quality of your
[4:29] output moving along the Persona
[4:31] component is basically who you want
[4:33] ChatGPT and Bard to be and the pro tip
[4:36] here is to think of someone you wish you
[4:38] had instant access to with the task
[4:40] you're facing if you enjoyed yourself
[4:42] working out that person might be a
[4:43] physical therapist with experience
[4:44] helping athletes recover if you're a job
[4:47] Seeker that person might be a recruiter
[4:48] or hiring manager if you're working on a
[4:51] creative brief that person might be a
[4:52] senior product marketing manager who's
[4:54] great at storytelling Pro tip you can
[4:56] also name specific individuals but I
[4:58] found the results to be good only when
[5:00] they're famous enough like Warren
[5:01] Buffett Steve Jobs Jeff Su
[5:04] by the way I just have to share this we
[5:06] have a team off-site with a superheroes
[5:08] theme so I asked ChatGPT to draft an
[5:10] email from Batman and it even included
[5:12] things like please let Alfred know and
[5:14] signed off as Your Dark Knight
[5:16] um so fictional characters work as well
[5:18] and I'm actually going to use this the
[5:19] fifth component format the pro tip here
[5:22] is to literally close your eyes and
[5:24] visualize how exactly you want the end
[5:26] result to look like a million likes on
[5:28] my thirst bomb Instagram photo damn it
[5:30] didn't work back to the user feedback
[5:32] example I don't want to read each
[5:34] sentence so I asked ChatGPT to take
[5:36] all the feedback and output a table with
[5:39] three headers the original feedback the
[5:41] team responsible for following up and
[5:43] priority and now I can copy this
[5:46] directly and paste it into a Google
[5:48] sheet sort by priority and filter by team
[5:51] other common formats include emails
[5:53] bullet points and code blocks but the
[5:55] one I found to be the most useful as a
[5:57] full-time working professional is
[5:58] paragraphs and markdown for example I
[6:01] just received a lengthy industry report
[6:03] for my director first give me the three
[6:04] key takeaways then summarize based on
[6:07] topic use H2 as section headers here is
[6:10] the report Pro tip whenever I use ChatGPT
[6:12] to proofread any document I specify
[6:15] that all changes need to be bolded so I
[6:17] can easily see exactly what has been
[6:20] changed let's quickly go through the
[6:21] last component tone before we put all
[6:23] this together in one example the good
[6:25] news is tone is easy to understand use a
[6:28] casual or formal tone of voice give me a
[6:31] witty output show enthusiasm sound
[6:34] pessimistic the bad news is we're
[6:37] usually not very good at recalling the
[6:39] thousands of potential adjectives and
[6:41] adverbs at a moment's notice so here's a
[6:43] pro tip tell ChatGPT the feeling
[6:46] you're going for for example I'm writing
[6:48] an email to a team I haven't worked with
[6:49] before and I want to be taken seriously
[6:51] without coming off as too stuck up and
[6:53] cringy can you please give me a list of
[6:56] five tone keywords I can include in a
[6:58] prompt for ChatGPT and look now in the
[7:00] actual prompt I can say use clear and
[7:02] concise language and write in a friendly
[7:04] get confident tone putting all this
[7:07] together let's look at this
[7:08] comprehensive prompt you are a senior
[7:10] product marketing manager at Apple
[7:12] Persona and you have just unveiled the
[7:14] latest Apple product in collaboration
[7:16] with Tesla the Apple car and received 12
[7:18] 000 pre-orders which is 200% higher than
[7:21] Target context write an email to your
[7:23] boss Tim Cookie sharing this positive
[7:25] news task and format the email should
[7:28] include at tl;dr too long didn't read
[7:30] section project background why this
[7:32] product came into existence business
[7:34] results section quantifiable business
[7:36] metrics and end with a section thanking
[7:39] the product and Engineering teams
[7:41] example structure use clear and concise
[7:44] language and write in a confident yet
[7:46] friendly tone tone note that if I had an
[7:49] existing email to reference I could
[7:51] delete the instructions around the
[7:53] structure and simply tell ChatGPT the email
[7:56] should follow the exact same format as a
[7:58] one I'll share below and paste the email
[8:00] from before by the way you can compare
[8:02] the output from this prompt to that of a
[8:04] simpler prompt I just launched a new
[8:06] product at the Apple car I received 12
[8:08] 000 pre-orders please write an email to my
[8:10] boss with this update there's a pretty
[8:12] big difference in terms of how generic
[8:14] and usable the end result is now that
[8:16] you know the basics of prompting my next
[8:18] video is going to take you from beginner
[8:19] to Pro so make sure you're subscribed
[8:21] for that check out my top five ChatGPT
[8:23] productivity tips for work see you on
[8:25] the next video in the meantime
[8:27] have a great one