Paste This Into Gemini And It Can Run Your Business

0:00 That's Gemini answering a question about my business. It's useless. And that's Gemini 18 hours later working before I got up. It's the same tool. Nothing about it changed. Is the same account.

0:14 And all I did was paste one document in. Six prompts. It took 40 minutes. And every one of them will go on screen so you can copy them and it will take you half the time. So here's where we

0:26 started yesterday. Somebody asked me to teach their team Gemini. I have built this setup dozens of times, just never in Gemini because I run my business on cloud. So, I know the

0:40 method. What I don't know is whether it would work the same way over there. So, I'm giving myself one evening and one of these six prompts is the only reason I would let this thing anywhere near my

0:52 inbox. It's also the one almost everybody leaves out. I'll tell you which one when we get to it. So, here's where we are starting. Completely new account and I'm going to ask Gemini,

1:05 what are the three things that I should focus on and why? And this is what I get. Generic advice. Review your goals, focus on high value tasks, and delegate where you can. I mean, that's not wrong.

1:19 It's just worthless. And it's worthless because Gemini has no idea who I am. And sure, what it says makes sense. Thanks, Gemini. But that's honestly what any LLM will give you when nobody has told it

1:34 anything about you. Now, you've probably been talking to a different AI for a year or two or three by now. Maybe Chad GPT, maybe Claude, maybe even Gemini, whichever one you ended up on. And that

1:46 one knows you. It knows what you sell, how you write, the thing you've been going around in circles on since March. All of that is sitting in another tab doing nothing for you over here. You

1:56 don't have to build it again. You can go and take it. There is a page in the settings called import memory to Gemini. Essentially, Gemini hands you a prompt to run in your other AI and it does the

2:07 job, but it's written about you as a person, and I want mine about the business. So, here's the one that I'm going to use. You and I have been working together for a while. I'm moving

2:18 my business setup to a different AI and I want to take everything you know about me with me. Please write me a single handover document in the second person addressed to the next assistant so it

2:29 can be pasted straight in as standing instructions. Please cover who I am and what my business actually does, who my customers are and what they come to me for, what I sell and what I refuse to

2:42 do, how I write and speak, including words I use and words I avoid, what I can honestly claim and what I must never promise, and anything you have learned about how I'd like to work. Please only

2:56 include things that I have actually told you or that you have picked up from our work together. If you're not sure about something, mark it as unconfirmed rather than stating it as a fact. That's it.

3:09 You can copy that off of your screen and paste that into whichever AI you've been using. And just look at what comes back. Mine already came back with a lot of information. my offers, how I write, the

3:21 claims I refused to make, two years of context that I did not have to sit down and write out. And there is a very important line in there that I think it's important to highlight. I told it

3:33 to flag anything it wasn't sure about instead of stating it. It's their nature to fill in gaps with something plausible if you let them. And a madeup fact about your own business is not what you want

3:45 turning in permanent memory. Now, copy the whole thing. Open a new Gemini tab and paste it in and put this underneath it. Everything above is what another AI knew about my business. Treat it as a

3:58 starting point, not as verified truth. Please read it and then interview me to fill in the gaps. Ask one question at a time and wait for my answer. Only ask about what is missing or marked

4:11 unconfirmed. Do not ask me anything the document already answers. Keep going until you can confidently write all of this. Who I serve and what they actually want. What I do and what I don't do. How

4:25 I want to sound in three words. What I can honestly claim and what I must never promise. Then give me one thing. Standing instructions under 200 words written in the second person for me to

4:39 just paste into personal context. That's it. Now, I want you to notice what I'm not asking there. I'm not asking Gemini what it knows about me because this account is a day old. You know, it knows

4:52 nothing. I'm asking what it's missing once it's read that document. So, it only asks the questions the other AI could not answer. If you've been working in Gemini before, you can also ask it to

5:03 enhance based on the memory and the information that it has saved about you. In my case, that would be nothing. Now, if those questions maybe stopped you in your process, you're in good company.

5:17 That is not a Gemini problem, and I'm going to come back to it near the end. Now, what comes out is 200 words of standing instructions, and those go into personal context. Whatever sits in that

5:29 box, Gemini will read before every chat you will ever open. You said it once, and then you stop thinking about it. Now, I want you to watch this. I'm going to start a brand new chat with nothing

5:40 in it. And I'm going to ask the same question that I asked at the beginning. And you see now it knows what I sell and I never told it. But read the answer properly though. It's about my business

5:55 now, which is a long way from where we were 3 minutes ago. But it's still soft. It's telling me what a business like mine should probably be working on this week because it knows about me, but it

6:06 doesn't have anything from me. 200 words of instructions is not the same thing as your actual work. So, that's the next prompt and I want to get into it. So, let's look at that answer again. It

6:17 says, "Focus on your highest value offer." Which one? It has no idea. I have several and it has never seen a single one of them. That is not something wrong with my prompt. It's a

6:29 supply problem because it only has my summary but none of my source materials. So, let's give it some. First, it needs somewhere to live. Right now, yours is probably in your inbox on your desktop

6:40 in a folder called stuff or work and maybe in a file that's named uh final_v3 actual or something like that. So, what you want to do is create three folders in your drive. what we sell, who we sell

7:00 to, and how we deliver. Three, not 10. Okay? If you can't instantly tell which folder a document belongs in, you're going to stop filling things within a week, and obviously the system breaks.

7:12 Now, Gemini Notebook. This is not a video about Gemini Notebook. You don't have to go in and work inside it. That is the bit that um will change how you use it because you want to build the

7:25 notebook once and then attach it to a normal Gemini chat. We can do it right here under more uploads. Your knowledge base sits in one place and you pull it into whatever you're doing. Okay. Now,

7:39 before I build it, there's something that we need to discuss. You should skip the notebook entirely and attach a drive folder straight to the chat. If you have three documents to your name, just do

7:50 that, okay? You don't have to over complicate things. I promise you it'll work fine. The notebook, however, makes a lot more sense when you have more than a handful documents, and that's for

8:02 three reasons. Number one, it's curated. A dry folder is everything you own, including last year's pricing, for example. A notebook is only the set you choose. Second, it sites. Every answer

8:16 tells you which document it came from, so you can click through and actually check. If you attach a raw folder, you'll get an answer with no receipts. And thirdly, it's one thing to attach

8:27 instead of five, which is the actual reason you will still be using it in a month. So, you can pick five to 10 sources. That's all. Maybe your business core page, your current offer, the

8:38 questions customers actually ask you, two pieces of work that you're proud of so it knows what good looks like coming from you specifically. I know the instinct here is to dump everything you

8:49 own into it, but please don't do that, okay? More sources make it worse. 10 pages of current accurate material will beat 100 pages of mixed material every single day because the messy 100 is

8:59 exactly where the wrong answers come from. Now with the notebook built, we can go back to the chat, attach it, and then paste this. What are the three most important things I should work on this

9:10 week and why? Use only the sources in the attached notebook. Cite the source for every factual claim. If the answer is not supported by those sources, say not found in the selected sources rather

9:23 than guessing. Please flag anything that looks outdated, inconsistent, or missing. You may have noticed that I provided four extra instructions through this prompt. Only these sources. This

9:35 way it stops reaching for the internet when my own document already has the answer. Number two, I asked it to site everything. This way I can check. This is the one that turns Gemini from

9:45 something you have to trust into something you can verify. Number three, I asked it to flag anything outdated or contradictory, which is more useful than it sounds because your documents

9:55 probably disagree with each other more than you think. And then the fourth one, I said, "Please mention not found instead of guessing." Now, watch this. I'm going to ask something that I know

10:06 for a fact isn't there. There. Not found in the selected sources. It just told me it didn't know. I need you to sit with that because it's really, really important. It's the first

10:18 genuinely trustworthy thing that this account has done. Almost every AI that you have used would have invented something right there and said it with complete confidence. This one looked

10:28 came up empty and then told me. So basically we asked the exact same question as before and now it is naming my own documents back at me. It's ranking these against what is actually

10:39 happening in my business this quarter instead of what a business like mine should theoretically be doing. So that is prompt two and we are two steps into something that can genuinely run. Now

10:50 there is one irritating thing left. I had to set up all of that to get an answer. Tomorrow if I start a new chat, I'm back to attaching the notebook, framing the question properly, reminding

11:02 it how I want things written, and so on and so forth. Wouldn't it be better if it just worked like that by default? Well, that's prompt three. A gem. A gem is basically a saved version of Gemini

11:15 that already has its instructions. We set it up once and every time we open it, it's already the thing we configured. That distinction is really the key. The notebook is what it knows.

11:28 The gem is how it behaves. Knowledge in one, behavior in the other. Okay, hardly anyone separates those two. And it's probably the reason most people's setup feels chaotic and like a mess. Now,

11:40 behavior is three things and almost everyone only does one of them. So, let me explain. It's where it gets the information, how it writes, and how it decides. And that last one is the one

11:51 people never think about. Every time you ask the LLM what matters most, it is making judgment calls on your behalf. Right now, it's making them based on what it reckons a sensible business

12:03 would do. It should be making them based on how you actually decide. So all three need to go in into our gym. So here's prompt number three. You are the business assistant for you can include

12:16 your business name. My customers are include your audience. What I sell is include your offer. My voice is include three words. What you use take facts about my business only from the notebook

12:30 attached to this gem and from my personal context. If something is not in there, please say so and ask me rather than answering from general knowledge. Please site which source you used for

12:41 every factual claim about my business. You may use the web for outside research, but label it clearly as outside information and never blend it into facts about my business. how you

12:52 write. For everything you give me, set it out as the goal, who it is for, the key message, the draft, and factual claim that I need to verify, and anything that needs my approval before

13:03 it goes anywhere. Never invent a price, a testimonial, a result, a guarantee, or any legal, financial, or medical claim. If you're missing a fact, ask me for it instead of filling the gap. How you

13:15 decide. When I ask what to work on or what matters most, rank by these rules in this order. And then you can include your own reasoning. What genuinely comes first in your business, what needs to

13:27 come second, what you deliberately push back and why. And then you continue by saying, always tell me what you move down the list and the reason. Flag anything you cannot rank because you're

13:40 missing information. Now, notice that I'm repeating the sourcing rule from prompt two. That one only ever applied to that one chat. In here, it becomes permanent. And that middle instruction

13:51 about the web is really important to spend a second on. A flat never searched the internet would be wrong because you do want it researching your market. The rule that actually makes it work for you

14:03 is that it can go outside. It just has to say so. And it never lets outside material next to your own facts as if it's the same thing. And then the last block, which is the one that I really

14:13 want you to look at, where things compete. Here's my order. Here's what I drop and here's why. And tell me when you can't rank something because you're missing information. That is my own

14:26 judgment, okay? Written down by me one time, but applied by it every time without me ever repeating it. You want to give it your notebook as its knowledge. And then you're done. So

14:38 let's ask the same question the third time. Right? First time asking and I didn't attach anything or remind it of anything. My structure, my language, but

14:51 look at the bottom of the answer because that's the new part. It's telling me what it pushed down the list and why. It's ranking with my logic instead of its own. So that's prompt number three.

15:02 You stop reexplaining yourself and it stops guessing at your priorities, which means it's now making judgment calls on your behalf. And that right there, and I mean this genuinely, is exactly the

15:14 moment that this gets dangerous. Two prompts from now, I'm going to show you why. So stick around. But for now, there is still one thing that it cannot do. It writes like me, but it has no idea what

15:25 our company looks like. And that is prompt four. Right now, every single time I want something made, I'm describing our brand from scratch. Our colors, our style, the thing we'd never

15:36 put on a page, every time from nothing. Now, I need to do a quick orientation first because there are three ways to make something in here. And a lot of people tend to mix these a lot. First of

15:47 all, it's image for the things you will use again and again. Video when you need a few seconds of motion, and canvas for when what you need isn't a picture at all. It's a document or a deck or a

15:59 page. Of course, there's music as well, but you can ignore it for now. We're running a business that does not have much to do with music. You don't need to master any of those today. What you need

16:09 is one more block in the gem we just built. So, it stops asking you what things should look like. So, here is prompt number four, and you want to add this to your gem underneath the other

16:20 three blocks. How things look, who should see themselves in our work, and what they should feel, visual personality in three words, colors we use, colors we never use, photography or

16:34 illustration style, the formats we need most, never include logos we don't own, invented dashboards, stock cliches, unreadable text, whatever the case may be. When something I am working on needs

16:47 a visual or a document, do not wait for me to ask. Produce a first version using whichever of image, video, or canvas fits the job. Tell me why you picked that one and leave clear space for

16:59 headline text. That's it. And I want you to notice where this is going. Okay? This is not going into personal context. That box gets read before every chat you open, including the ones that have

17:10 nothing to do with work. So, it stays small. This is how the business assistant behaves. So, it lives in the gem with everything else. That last paragraph is very very important. That

17:22 is the instruction. Now let's run the same question fourth time and read the top of it first because the answer has not really changed. The same priorities, same order, same reasoning

17:36 as last time. It's what sits underneath that that's new. The first job on the list came back with a draft already made. On brand, right format, and it's telling me why it chose that format over

17:48 the other two. I never asked for any of it. It read its own plan, worked out what the first job actually needed, and went and started it. So, it plans, it writes, and now it produces without

17:59 being asked. That's four prompts. And this is exactly where most people stop. Maybe connect it to their inbox and let it run. I also connect it, just not before prompt five. So, I want you to

18:10 look at what it can do now. Okay? It knows the business. It writes like me. It decides like me. It makes things on its own without being asked. Notice what it still can't touch, though. Basically,

18:22 nothing. No inbox, no calendar, no contacts. It's been sealed in a box this whole time. And we need to change that in a few minutes. And this is the prompt that I need to write first because

18:34 here's what happens when you skip it. You're going to connect it to your Gmail. Maybe you connect it to the calendar. You let it get on with things and that's fine. It's fine right up

18:43 until the morning it drafts a reply to a customer or works out a price from a document that's 18 months old and it sends it confidently before you're even awake. It will not check with you

18:54 because nobody ever told it. So, we need prompt five that you want to add to your gem underneath the other blocks. what you do alone and where you stop on your own without asking me. You can do

19:08 research, summarize, sort, categorize, draft, prepare a first version, and set reminders for me only. Stop and get my approval before any of these. Sending anything to a customer, publishing

19:20 anything publicly, quoting a price, changing a customer record, committing money, agreeing to a deadline, or making any legal financial or medical or employment claims. If a job crosses that

19:33 line halfway through, do the part you're allowed to do, stop at the line and tell me you have stopped and why. When you give me a plan, mark every item that needs my approval before it can move.

19:46 Now, I want you to read that second list because it's the one people argue with me a lot. Sending anything to a customer, publishing, quoting prices, changing a customer record, spending

19:56 money, agreeing on a deadline, anything legal or financial. And usually the push back tends to be the same. That's most of the useful stuff. You're crippling it. No, everything on the first list

20:06 still happens on its own all day while I'm nowhere near it. It researches, it sorts, it drafts, it prepares, it gets things ready to go. What changes is that the last inch, the irreversible inch,

20:19 stays with me. And that is exactly what lets me walk away from it. So, I'm going to go ahead and ask the same question the fifth time. Now look at what it's added. Two of

20:32 those three priorities have come back marked because they need my sign off before anything moves. And it's telling me which specific part needs it. It's still thinking entirely on its own. It

20:43 just stops before it does anything that I can't undo. So now you have five prompts. We have one to go. And now I can let it out of the box. So moving to the sixth one because everything up to

20:56 here still needs me to show up and ask. I need to open Gemini. I need to ask. It answers brilliantly. And then I shut the laptop and nothing happens until I come back. A system doesn't need me to

21:09 remember it. This still does. And the first week I get busy, it will probably stop. So we need to take care of some scheduled actions. Now, a scheduled action is a standing instruction with a

21:21 clock attached. You write the job once, you tell it how often and what time, and from then on, Gemini goes and does it without you and has the answer waiting. Now, one quick thing up front before you

21:32 get attached to the idea on my account. This sits behind a paid plan, so please check whether you've got it. The five prompts we've already done all work without it. This is the only one that

21:44 doesn't. If you have it, this is the best 15 minutes of the whole setup. If you don't, you still have a business assistant that knows your work and follows your rules. You just open it

21:53 yourself. Now, the second thing, it does not email you. Okay? Even if you connect it to your Gmail account, the answer will be sitting in your Gemini chat when you open it. That box that says deliver

22:05 by is the time, not the destination. There are templates in here, by the way. You can use a news digest, a daily motivational quote. We're going to do something considerably more useful. So,

22:16 first connections because it can't tell you what matters today if it can't see today. So, you want to connect an app only if you can say the reason out loud in one sentence. Gmail so it can see

22:27 what came in overnight. That's fine. Calendar so it knows what today actually looks like. It's fine. If you can't finish that sentence, don't connect it. And trust me, you're not going to miss

22:36 anything. I'm going to connect two. That's all. And I'm only comfortable doing that because of the prompt that I wrote a few minutes ago. It can read all of this. Now, it still can't send a

22:46 single thing without me. Now, the last prompt. This one goes in the instructions box. Use my And here, add your gem name, and follow all of its instructions. Look at my calendar for

22:57 today, my overdue tasks, and my most important unread emails. Then answer this. What are the three most important things I should work on this week, and why? Please rank them by rule one, rule

23:11 two, whatever you need and push down based on rule three. Give the reason for each one and name where you got it from. Mark anything that needs my approval before it can move. Do not invent

23:23 anything. If something is missing, please say so. Now, you can give this a name. You can set it to daily. You can pick your time. And that first line points it at the gem. So, everything

23:33 we've spent the last half an hour building will come with it. It does not start from nothing in the morning. Now, one honest detail before we finish. These get prepared up to an hour before

23:44 your delivery time using whatever information exists at that point. So, an 8:00 briefing might have been built at 7. For a morning plan, that's completely fine, but just know it works that way

23:57 and pick your time and your job to suit it. Now, let's read that line again. What are the three most important things that I should work on this week and why? That is the same question that I have

24:08 been asking five times today, word for word, after every single prompt. Now, this is the sixth one and it's the first one that I'm not going to ask. It took us about 40 minutes. Now, it's time to

24:20 close the laptop and I'll see you in the morning. Good morning everyone. It is now the morning after. I have not touched this. So, let's open the computer and see what

24:36 happened. Okay, we've got three things ranked. Number one with the reason and the document that reason came from. Number two, same. And number three, it's pushed down the list and it's told me

24:49 why it pushed it down. Two of them have come back flagged. I think that's because those need me before anything moves. And down here at the bottom, it has found something it couldn't rank

25:00 because it's missing a piece of information. And it's asked me for it rather than guessing. And I was asleep. Okay. Nobody asked it for any of this. It looked at my calendar, went through

25:12 what came in overnight, read its own instructions, applied my rules, and had this before I got up. And the thing I actually wanted to know last night, whether any of this would work the same

25:23 way over here, clearly it does. Six moves in a different tool one evening. Now, in a minute, I'm going to put last night's answer and this morning's side by side because the distance between

25:35 them is honestly ridiculous. But before that, I need to be honest. 40 minutes was fast. And it was fast for a very specific reason. When prompt one asked me who I serve, what I do, what I refuse

25:49 to do, and what I'd never promise, I actually already had those answers ready. That's the only reason the other five worked. If you got to prompt one and went blank on those questions, that

26:00 is not a Gemini problem and no tool on Earth can fix it. It's just the first time anybody's made you write it down. That page is one of the first things that we build inside Aros. It's the four

26:12 module course in the AI founders community. And module one is exactly about this. getting what you actually do out of your head and onto a page with a virtual room full of people doing it at

26:23 the same time as you. The link is here as well as in the description if you'd like to join. And if you'd rather do it on your own, you now have all six prompts. Nothing in this video is behind

26:34 anything. You can just join the free community if you want to copy paste them, but you can just screenshot from the video itself. Now let me show you last night's answer. This morning's same

26:48 question, same tool, same account, six prompts and one evening in between. Gemini was actually not underpowered. Nobody had onboarded it. And the thing that actually changed isn't in Gemini at

27:01 all. It's that my business now exists somewhere outside my own head that is not clawed. Those six prompts made me write down what I sell, who I serve, how I decide, and what I'd never let go out

27:14 with my name on it. The scheduling was actually the easy part. And if you'd rather do this all in Claude, I've actually already built that one. You can watch that one next, and it's the

27:26 version that I actually run my business on. Thank you so so much for watching. Like this video if you did. Be sure to subscribe if you haven't done so. Share this with anyone in your circle of

27:36 friends or family or co-workers who might find this useful. Maybe they already run in the Google environment and Gemini is easier for them to operate in. Thank you again. Watch this video

27:50 next and I'll see you there. Bye.