Google Prompting Guide 101 - As a Podcast

Jeremy Ford
479 followers
160 Video Views·Sep 21, 2025
I was reading Google's prompting guide, and decided to use Notebook LM to make it into a podcast.
This is a special message for just those watching this on Ganjing World.
I realize a lot of us are here on Ganjing World because we want to get away from a lot of the
shifts that are happening in society and a lot of the progress, so to speak. But I'm making content
about AI to educate people and show people that even though it does have its concerns, that it
can be useful. And I want to encourage you to watch my other video that I posted directly to
Ganjing World titled, A Special Message to Ganjing World. I'm trying to encourage people to learn
about AI, not to use it necessarily, but because when you understand it better, a lot of those
fears and concerns can go away. And I explain it well in that video. So even though this video is
about AI and uses AI, I want to encourage people to go look at that other video to understand my
motives for focusing specifically on this topic. Since it's a Saturday, I was upping my skills for
prompt engineering and I was reviewing Google's prompt engineering guide that they released,
I think it was a few months ago, and it's 68 pages. I've seen it before, but I decided today
just to put it into Notebook LM and I made a 22-minute podcast for it. And since I already
went through the trouble of doing that, I thought I would save you the trouble and just upload it to
YouTube. So here is that podcast. Welcome to the Deep Dive, where we really try to turn that feeling
of information overload into, well, actionable knowledge. Yeah, useful stuff. Exactly. And today
we are focusing on what feels like the ultimate shortcut to productivity getting good at generative
AI. Forget the really generic advice. We're aiming to cut through all that noise and focus on maybe
the single most actionable metric we've seen in AI prompting. 21 words. That's right. And this
Deep Dive is really for the learner out there who gets that AI isn't a fringe tool anymore.
No, not at all. It's becoming core to how we work professionally. We're talking about
really mastering the conversation with AI assistants like Gemini. And specifically Gemini
because it's built right into the tools many of us use every single day in Google Workspace.
Exactly. That tight integration is, I think, the key thing here. Yeah, because we aren't talking
about popping open a separate window or app. Nope. This generative AI is actually embedded right
there. Gmail Docs, Drive Sheets, Meet Slides. All of them. Which means you can kick off a task
right inside a familiar document, say, and use the AI without breaking your stride,
without losing your flow. Which sounds much more efficient. And this integration,
it lets you work directly with your own information, your personal knowledge base,
the file sitting in your drive. Ah, okay. And crucially, doing that while keeping that
enterprise-grade privacy and security that Google always emphasizes.
That's critical for businesses. Absolutely. So our mission today is pretty straightforward,
actually. We want to distill the essential skills you need to write genuinely effective prompts.
Making AI feel less like a tool you command and more like a real-time collaborator.
Precisely. Boosting productivity, creativity, all while feeling secure.
Okay, let's unpack this 21 words thing straight away. Because you said the average person is
using less than nine words. Yeah, that's the surprising data point. Less than nine words
on average. But the best results, the really fruitful prompts, they average around 21 words.
So what's happening in those extra words? What's in that optimal prompt?
Well, it boils down to specificity. Those short prompts, they're basically like shouting a command
into the void. Right. Summarize this.
Exactly. Whereas the 21 word prompts, they give the AI enough relevant context to act like a
genuine specialized collaborator. Okay.
That higher word count isn't just padding. It comes from applying a clear structure.
We're calling that structure the power prompt. I like that. The power prompt. So the anatomy of it,
the research you mentioned points to four main areas or components.
That's right. Four key components. And you don't always need every single one,
but using them drastically improves the outcome. Correct. They are persona,
task, context, and format. Okay. Let's quickly define those. First up, persona.
This is about telling the AI who it should be. Yeah. Who should it act as? Giving the AI a role
like, say, a program manager or a PR executive, maybe even a specialist in a certain industry.
Ah, like a financial analyst persona. Exactly. It instantly sets the kind of tone you expect,
the knowledge base it should draw on, and the scope of the response.
Makes sense. Second. Second. And this one is absolutely the most critical element,
the task. The verb. The verb. The command. You have to tell it what to do. Summarize, draft,
analyze, generate, compare. Without that clear action, it just can't give you anything useful
back. Got it. Task is essential. Third is context. Context. This is all that necessary background
information. What industry are we talking about? Are there specific documents or
goals the AI needs to know about? So this is the why behind the task. Precisely.
Feeding it the relevant background. And lastly, format. How you want the answer delivered.
Exactly. The output structure. Should it be bullet points only? A concise table,
maybe a formal tone versus something more casual. You get to control the final packaging.
Okay. Four components. Persona, task, context, format. But I have to ask this because I can hear
people thinking it. Go ahead. If I'm just asking it to summarize, say, an internal memo for myself,
do I really need all four? Isn't adding persona and context just making the prompt longer than
it needs to be? That's a really practical question. Yeah, you're right. Sometimes you
don't need all four explicitly spelled out. Okay. But here's the thing. Using even a hint
of persona and context helps make sure the output isn't just generic fluff. Oh, I. For instance,
asking summarize this memo, fine, it'll do it. But asking as a chief legal officer,
summarize this memo, focusing only on potential compliance risks and deliver the output in a
numbered list. Okay. Yeah. That's much more targeted. It's far faster and way more valuable
in the end. It turns a generic summary into specific actionable intelligence. It's about
getting the most out of those words. That makes perfect sense. It's about maximizing the value
you get back. So let's bring all four together. Give us that single comprehensive example prompt
using everything. Okay. A solid four-part power prompt might look something like this.
You are a program manager in the, let's say, fintech industry.
Draft an executive summary email to the SVP of product based on details about the Q3 launch
program docs you've maybe linked or described. Limit the final result to three bullet points.
See, that leaves very little room for error or misinterpretation. It's super clear.
Exactly. Nothing left to chance. Okay. So we've mailed down the structure,
the anatomy of this power prompt, but prompting isn't usually a one-shot deal, is it? It's more
of an interaction. Definitely. So once we send that first prompt, the real productivity boost
often comes from refining it, iterating on it. How do we keep that flow going without writing a
whole new 21-word prompt every single time? Right. And that's where those six quick tips
really come into play. And it starts with a mindset shift, actually.
Okay. Thinking of it less like a search box and more like a dialogue, a conversation.
Oh, okay. Which leads to the first tip. Use natural language. Just write like you talk.
Complete thoughts, full sentences. The AI actually understands nuance and context better
when you converse with it. Right. Ditching the old
habit of just typing keywords. Exactly. No more keyword soup.
Okay. So natural language is the starting point. How do we make sure that conversation
stays on track and productive? Well, that brings us to the second and third
tips, which kind of go hand in hand. Be specific and iterative, but also be concise.
Specific but concise. How does that work? You need to clearly state the action,
summarize, write, change the tone, whatever it is. That's the specific part.
But you also want to avoid unnecessary targon or complexity. Be brief when you can be,
but don't sacrifice the critical specifics. Got it. Clear action, no fluff. And if that
first result isn't quite right, maybe the tone's off or it missed something important.
Yeah, that happens. And that's where the conversational aspect really shines.
That's the fourth tip. Make it a conversation. Don't just give up on the first try.
Absolutely not. Never settle for the first answer if it's not right.
Use short, quick follow-up prompts. Things like,
make that sound more diplomatic, or could you expand on the second bullet point?
Just little nudges. Exactly. Refining is almost
always quicker than starting completely from scratch.
Okay. That makes sense. And the fifth tip is the one that feels like a real game changer,
especially for people working inside Google Workspace.
Oh yeah. Use your documents. This ability to pull context directly
from your own files in Google Drive? It's genuinely transformative. You can
personalize the output based on your reports, your memos, your data,
just by using the at file name tag right there in the prompt.
So the AI isn't just scraping the public web? No, it's drawing from your specific,
often confidential organizational knowledge that gives you a level of personalization
and importantly, security that's just, well, unmatched by generic tools.
Huge advantage. Okay. And the last tip for those using the more advanced chat experience.
Right. For Gemini advanced users, there's this cool meta tip, make Gemini your prompt editor.
How does that work? You can literally start a prompt by saying,
make this a power prompt. And then you paste your original, maybe shorter prompt text here.
Oh, wow. And Gemini will actually analyze your attempt.
It'll give you suggestions on how to improve the structure. Maybe it'll say,
consider adding a persona, or you should specify a format.
So it teaches you how to write better prompts while you're using it.
Exactly. It sort of automates the learning curve, which is pretty neat.
That's a really powerful set of tools and tips. Okay. Let's make this concrete.
Let's see how this applies across different jobs, different professional roles.
Starting with say, administrative support efficiency is absolutely key there.
For sure. Imagine an admin task with planning a multi-day team offsite meeting. Big job.
Yeah. Lots of moving parts.
So they could start using that power prompt framework right away.
I am an executive administrator. Plan activities for each day that include
team bonding activities and time for deeper strategic work. Create a sample agenda for me.
Okay. That gives them a starting point, a skeleton agenda.
Right. But it's probably too generic at first. So the admin needs specifics.
They iterate immediately with a follow-up.
Like asking for activity ideas.
Exactly. Something like, suggest three different icebreaker activities that
encourage people to learn about their teammates' preferred working styles, strengths, and goals.
Make sure the icebreaker ideas are engaging and can be completed by a group of 25 people
in 30 minutes or less. See how specific that is.
Very specific. Gets them concrete options.
Right. And once they have that detailed content,
the agenda items, the icebreaker's final step might be formatting.
Make it look clean.
Yeah. They prompt again. Organize this agenda in a table format.
Include one of your suggested icebreakers for each day. And boom, they've gone from a blank page
to a perfectly formatted, customized agenda, maybe in just three conversational steps.
That's a fantastic demonstration. And the key takeaway there, I think,
is that no single prompt has to be absolutely perfect on the first try.
Not at all.
It's really about that back and forth, that iteration.
Using the AI to refine, to reformat quickly until you land on exactly what you need.
Couldn't have said it better. Okay, let's shift gears. How about communications?
Think about a PR manager who needs to draft a press release after, say, a company acquisition.
High pressure, tight deadline usually. And accuracy is paramount, especially with quotes.
Absolutely. So the power prompt here would likely focus heavily on the task, yes,
but also that crucial file immigration we talked about.
Using the at symbol.
Right. I'm a PR manager. I need to create a press release with a catchy title,
include quotes from at VIP quotes acquisition document.
And that document is just sitting in their Google Drive.
Exactly. And that's the real leverage. The AI isn't just spitting out generic PR language.
It instantly pulls the necessary executive quotes, maybe key context points
from that separate pre-approved document.
Directly into the new press release draft they're working on in docs.
Right there in docs. Imagine the time saved. No more hunting through folders,
copying, pasting, worrying if you grabbed the latest version.
Yeah, that's huge. It's like the AI is accessing the company's specific memory on demand.
That's a great way to put it. Okay, next up, customer service.
This role demands both empathy and often really fast problem solving.
Definitely. So imagine a scenario. A customer service rep gets an email about damaged goods.
The customer is upset and they've requested a non-standard solution
like expedited shipping, which the company maybe can't offer.
Okay. Tricky situation. So the initial prompt might use the persona element and focus on
getting the tone right. Help me craft an empathetic email response.
Setting the emotional tone.
Exactly. Include a paragraph that acknowledges their frustration and maybe three bullet points
with potential resolutions. So the AI helps draft that empathetic core message.
But the rep still has the problem. They can't offer that specific
expedited shipping the customer asked for.
Right. So instead of the rep struggling to write a polite no
and come up with policy compliant alternatives from scratch.
They iterate.
They iterate. They use a brainstorming follow-up prompt. Suggests 10 alternative
options in place of expirated shipping to resolve the customer's frustration
about receiving the damaged package.
Ah, so it generates options for them.
Instantly. And suddenly the AI might list things like offering a full refund,
store credit, maybe covering repair costs, a discount on the next order.
Wow. Okay.
Rapid brainstorming of viable policy compliant solutions.
This frees up the human agent to focus on the actual customer interaction,
the emotional connection, rather than getting bogged down in drafting alternative offers.
That's a great use case. Shifting again,
what about executives? Say a COO preparing for a potentially tough quarterly town hall meeting.
Yeah. High scapes communication. Here the use case shifts a bit
away from pure content creation towards more strategic preparation.
How so?
This is where something like Gemini Advanced can really shine,
helping anticipate potential issues. The executive could prompt it to think like the audience.
Like predicting questions.
Exactly. Something like, help me write challenging questions that
employees may ask at the upcoming town hall about,
and maybe they link to the internal company announcement URL.
Okay.
Then generate potential answers for each question that use a confident, but firm tone.
That is brilliant. So it's not just generating text, it's simulating the Q&A.
It's helping them anticipate and prepare for a critical communications scenario.
It gives the executive a strategic edge, allowing them time to think through and prepare strong,
confident responses to potentially difficult questions before they're on the spot.
Huge value in sensitive situations. Okay. One more. Let's bring it down to the small
business owner. Maybe a spa owner trying to choose between two cleaning service proposals.
Right. Juggling everything, finance, operations. They need quick decision-making tools,
especially when faced with dense documents like proposals.
Lots of details buried in there.
For sure. So their prompt would likely rely heavily on that file integration for comparison.
Using the ad again.
Yep. I'm a business owner and I'm trying to determine the right cleaning vendor
using at company A proposal and at company B proposal.
Referencing both files directly.
Create a comparison table between the two companies' proposals.
Simple request.
Simple request. Powerful result. The AI then reads both those dense service agreements,
pulls out the key details, pricing structures, standard rates,
minimum hours, cancellation policies, whatever's relevant.
And puts it side by side.
Exactly. Synthesizes it all into a simple, easy-to-scan comparison table.
It transforms complex document analysis, which could take hours,
into a straightforward decision-making tool in potentially seconds.
That's incredibly practical. These examples,
they really do show the sheer breadth of how Gemini can be applied, don't they?
Absolutely. It operates across pretty much every professional function imaginable.
Whether you're, I don't know, tracking IT assets in Sheets.
Or maybe developing detailed customer journey maps in Docs.
Or even generating unique visuals for a marketing presentation in Slides.
It seems to come back to leveraging that power prompt structure we talked about, persona,
task, context, format, and embracing the fluidity of the conversation, the iteration.
That's the core of it. Delegating the tedious stuff,
the time-consuming analysis or drafting, to the AI.
So if the main lesson today is that prompting is a skill,
a very accessible skill that really anyone can learn.
Yeah, I believe it is.
Then the potential productivity rewards are just enormous.
But there's always a but, isn't there?
One final, really crucial reminder to actually realize those gains.
Yes, absolutely critical.
And that is, generative AI is incredibly powerful. It's your assistant.
But the final output, it's still yours. It has your name on it, metaphorically speaking.
You own it.
Which means you must always review the result.
Check it for clarity, for relevance, and maybe most importantly, for accuracy.
Before you hit send, before you act on it.
Always. The AI does the heavy lifting, the first draft, the synthesis.
But the human absolutely remains the editor-in-chief.
You have the final say.
Wise words.
Okay, so we've covered the four key components of a power prompt.
Those six quick tips for making it a productive conversation.
And we've seen some really practical examples
of how different roles can use this technology right within their everyday apps.
Yeah, we've covered a lot of ground.
Hopefully, it demystifies prompting a bit.
I think it does.
So now, a final thought for you, the listener, to mull over as you finish this deep dive.
Okay.
If prompting really is a skill you can start mastering today, right now,
what's the most complex, maybe multi-step workflow in your job?
You know, that project you kind of drag starting because it feels overwhelming.
We all have those.
Could you now see a way to break that down?
Delegate parts of it, the research, the drafting, the analysis to your AI assistant
using these techniques.
Freeing you up.
Freeing you up to focus on the truly human parts of your work.
The strategy, the relationships, the final decisions.
Think about that next potential breakthrough project
and how AI might help you tackle it differently.
