Build the Business Brain

[00:00:00] Building the business brain.
[00:00:04] Now that we have Claude locally installed along with Cursor on our computer,
[00:00:09] we can finally get to building the business brain.
[00:00:12] And guys, this is one of the modules that I am the most excited
[00:00:16] for. By the end of the lesson, we will have our own business brain,
[00:00:20] our company on paper, even if the company is just ourselves individually.
[00:00:24] The first client profile inside of it and build it from four
[00:00:28] call transcripts and two messy intake notes without inventing anything.
[00:00:33] First of all, what we need to have is we want to have this
[00:00:36] page open in your browser. So just keep this along as we're going over
[00:00:39] the lesson. Number two, we want to have Cursor open with the relay business
[00:00:44] brain we want open. So the way we do that is I have Cursor
[00:00:48] open. And what you do is you open projects and you simply
[00:00:53] select the zip file and the folder that you downloaded onto your computer.
[00:00:57] So I already have it open over here and I simply just click on
[00:01:00] it. As we can see, this is the Cursor terminal and
[00:01:05] this is the system and the interface that we're going to be using to
[00:01:08] build the business brain and interact with the AI.
[00:01:10] And finally,
[00:01:13] what we want to be doing
[00:01:18] is using Cloud Code in
[00:01:23] Cursor's built-in terminal.
[00:01:32] Open the terminal panel and type Cloud Code. So,
[00:01:35] simply said, we just come here and we type Cloud.
[00:01:38] We click on Yes, I trust this folder. And now we have.
[00:01:42] We can do slash model. We can pick the different models that we want
[00:01:44] to select. We can do slash effort.
[00:01:47] We can select the effort level of the AI and whatnot.
[00:01:57] So,
[00:02:03] before we get in, before we get into it,
[00:02:05] I want to, before we get into it,
[00:02:08] I want to explain why exactly do we have the business brain over here.
[00:02:13] The reason being is that most people's AI,
[00:02:15] it has no memory. Every chat starts from zero where you're re-explaining your business
[00:02:20] for the hundredth time. And the answer may sound smart,
[00:02:23] but that's not grounded in anything. In facts,
[00:02:26] in data, your CRM data,
[00:02:28] all the calls you have, your employees, the different clients you have.
[00:02:32] But, a business brain over here, that fixes that.
[00:02:35] A folder of plain files that hold everything that the AI needs to know
[00:02:39] about the business. And simply, the AI just parses through all of these files,
[00:02:43] reads them, Thank you. It does the work and shows you where every single
[00:02:47] fact came from. And by the end of this video,
[00:02:50] we want to have our own brain built with our first client inside of
[00:02:53] it. Every lesson after this one builds on this exact folder.
[00:02:57] And this is the core foundation that the whole course is going to be
[00:03:01] standing on. So let's open back this document over here.
[00:03:04] What we want to do is, in Cursor's file,
[00:03:06] guys, there's going to be a starter document.
[00:03:09] Starter.md, a business brain template, and a company 1 medspa.
[00:03:13] So let's take a look at these files. Let's click on starter.md.
[00:03:17] A simple trick when you're reading .md files.
[00:03:20] .md stands for markdown. That is simply the format that is the most suited
[00:03:24] for AI to go over information. This is what it looks like on Cursor.
[00:03:28] But if we come into preview, we can see that it becomes a lot
[00:03:31] more legible for us as humans to actually use. To go through the information.
[00:03:34] So the starter.md is simply the file that contains all of the information about
[00:03:39] how to set up Claude on your Mac and whatnot. Then what we have
[00:03:41] is the business brain template where I'm going to close this and zoom in
[00:03:43] a little bit. Let's take a
[00:03:57] look at this. What we have are calls,
[00:03:59] clients, core, data, output,
[00:04:02] and rules. If we open calls,
[00:04:05] we can see the .readme file and let's click on the preview and let's
[00:04:08] go over it. So this is the format guys that we want to be
[00:04:11] instilling information into the AI.
[00:04:13] We need to give it clear rules. We need to give it guardrails about
[00:04:17] how you need to go over this information and as a result.
[00:04:21] Of updating or scanning a piece of information.
[00:04:24] What is the downstream effect of other files that get updated with it?
[00:04:28] So we have calls and we go over this. We have clients over here.
[00:04:32] We have core fundamental information about our own business.
[00:04:35] Correct? What is the offer that we have? Who are the customers?
[00:04:38] What is my personal goal for this business? Because then this allows the AI
[00:04:42] to truly be a strategic business advisor.
[00:04:45] Rather than just a generic AI who's just guessing what we're trying to do
[00:04:49] over here. The voice. We have data,
[00:04:52] we have output, and we have rules.
[00:04:55] So first, let's go over the sample company data in,
[00:05:00] So first, let's go over the company one, MedSpa,
[00:05:02] the sample data information. Let's open the readme information.
[00:05:06] Let's go over to preview. So this is an entirely fictional business.
[00:05:09] MedSpa created for this course, and every company detail,
[00:05:12] person, price, call, and transaction is invented.
[00:05:57] go over Company 1, MedSpa. So,
[00:05:59] this is a sample piece of data that essentially replicates
[00:06:04] what a real client environment looks like.
[00:06:07] What we have over here, guys, are the raw intakes,
[00:06:10] the data, the calls, and a real-life demonstration
[00:06:15] of how, when you're auditing, working,
[00:06:17] and scoping out systems with companies. You are providing.
[00:06:21] information in the form of. Let's take a look at calls.
[00:06:24] So, what we have over here are the raw call transcripts.
[00:06:27] For us, going from discovery, we can take a look over here,
[00:06:31] all the way to data walkthrough, when we're working with the technical team in
[00:06:36] the company, the owner consultation,
[00:06:38] when we're potentially finally talking to the owner to close in the deal,
[00:06:42] and the front desk process. So, once the invoice has been sent,
[00:06:46] and the deal has been closed, and we begin delivery,
[00:06:49] we start scoping out exactly what is the front desk process,
[00:06:52] and how exactly are these people working and moving,
[00:06:56] and how exactly are they working together.
[00:06:59] Then what we have over here are the raw intakes.
[00:07:09] Then what we have over here are the raw intakes,
[00:07:14] and as you can see guys, so first of all, let's
[00:07:33] So first of all, let's go over company 1 med spa,
[00:07:37] which is a fictional client.
[00:07:47] So first of all, let's go over company 1 med spa,
[00:07:49] which is a fictional database of what a real life environment when
[00:07:54] you're working with a client looks like, the exact type of information you need
[00:07:58] to work with. So first let's go over
[00:08:03] company 1 med spa,
[00:08:07] which is a fictional
[00:08:21] data walkthrough to the owner consultation to us understanding what exactly
[00:08:26] is the front desk process for this company.
[00:08:29] As we can see over here, Citrine Val Aesthetics,
[00:08:32] every company, every person mentioned.
[00:09:10] Let's begin this
[00:09:15] by going over Company
[00:09:22] 1 MedSpa.
[00:09:25] So what we have over here is Citrine Val Aesthetics.
[00:09:28] This is a sample dataset and a fictional,
[00:09:32] So first of- So
[00:09:37] first of all, let's go over Company 1 MedSpa.
[00:09:40] If we take a look at the readme.md file, we can see that Citrine
[00:09:42] Val Aesthetics is a, completely...
[00:09:46] is a... So first of all,
[00:09:51] let's go over Company 1
[00:09:55] MedSpa. Give these two, the readme.md file and open,
[00:09:59] and as we go through it, we'll see that Citrine Val Aesthetics is a
[00:10:02] sample dataset that we've created in order to really emulate and simulate what
[00:10:07] an environment of working with a client looks like.
[00:10:10] What we have over here is we have the calls,
[00:10:13] every single, the discovery call, the data walkthrough,
[00:10:16] the owner consultation, and the front desk process that we had with them.
[00:10:20] Score. Open out the system, closing the deal,
[00:10:23] the data that the client in this case has provided to us,
[00:10:26] the CRM data, the metrics and the ad set data,
[00:10:30] then the raw intake. So as we can see guys,
[00:10:33] a lot of the data over here is really messy.
[00:10:37] I'll really encourage all of you to give it an open and go through
[00:10:40] each one of these images. We can see we have an image of an
[00:10:43] ads manager. A booking calendar over here.
[00:10:47] Then we have intakes note.md where let's convert this to preview and we can
[00:10:51] see it is really, really messy data.
[00:10:53] There's no formatting. We have random dates,
[00:10:56] the go high level screen share, couldn't find the share button,
[00:10:59] the marketing calls, the front desk.
[00:11:02] Then we have the invoice export where let's click into this and what What
[00:11:07] we can do is, when you have a PDF, you can simply,
[00:11:09] Click on it and say open in browser.
[00:11:12] Right. This is what the invoicing looks like.
[00:11:15] Then we have the screenshots for the ads.
[00:11:17] Once again, if it's an HTML as well, you can click on open in
[00:11:20] browser and you can take a look at it. So what we have over
[00:11:23] here, guys, is this is what it actually looks like in week one.
[00:11:26] When you're working with a client, four call transcripts,
[00:11:30] a scratch note file. You and a voice memo the owner recorded in a
[00:11:34] grocery store parking lot and close. Nobody really hands you a spec.
[00:11:38] There's no one who's going to be handing you clean data.
[00:11:41] And this is something
[00:11:55] Where if we open the ownervoicememo.md,
[00:11:59] the owner has essentially sent over a voice transcript.
[00:12:03] It could be over WhatsApp, on Slack, it could be on iMessage, and they
[00:12:05] have just screen recorded themselves. This is exactly what happened to me.
[00:12:08] I mean, this is exactly what I've seen happen.
[00:12:30] And this is a very common scenario that we will be noticing.
[00:12:34] If we take a look at this, I'm doing the voice note because I
[00:12:37] sat down here to write this, you're never getting it.
[00:12:40] I'm in the grocery store parking lot, and if this sounds familiar to you,
[00:12:42] it's because I'm in the grocery lot.
[00:12:48] I'm in the grocery store parking
[00:13:06] common and that is our entire purpose to stop them from running a business
[00:13:10] that is running in their head, systematize everything,
[00:13:13] leverage the intelligence of AI on top of it,
[00:13:16] and voila. The value that these guys get,
[00:13:19] different business owners, is tremendous.
[00:13:23] We spent 1900, got 23 leads in the report,
[00:13:26] and I remember it as a two person, as two people,
[00:13:30] because this guy used to what it felt like in the rooms, FET-like,
[00:13:32] they're going completely based on vibes. Also,
[00:13:35] the Google numbers may not be current, Lexie says that Connector needed attention before
[00:13:39] Brin started the review process, and as we can see,
[00:13:43] it's complete chaos. But,
[00:13:53] and this available. And this is a very common and recurring theme,
[00:13:57] guys. Nobody is really ever gonna hand you a spec.
[00:14:00] Nobody hands us clean data, and in the voice memo she says herself,
[00:14:04] I send the transaction report, the screenshots,
[00:14:07] the notes, and that is pretty much the entire business.
[00:14:18] If
[00:14:27] your system needs clean inputs,
[00:14:30] you don't have a system. The whole point of what we're building with the
[00:14:34] business brain, leveraging the intelligence of AI on top of it,
[00:14:38] is that it eats up this entire mess and produces something incredibly bad.
[00:14:42] And
[00:14:48] one thing about these transcripts guys,
[00:14:50] the consultant you'll see in them, that is the operator's seat,
[00:14:54] over here. So that is how you want to visualize yourself when you're asking
[00:14:58] questions, scoping out these exact calls.
[00:15:00] Over here, Jordan Boss, that is the person who we want to put ourselves
[00:15:04] in this person's shoes. where he's out there.
[00:15:06] Asking the right questions to the business owner,
[00:15:10] scoping it out, trying to understand the workflow,
[00:15:12] and really understand where lies the opportunity for creating significant value
[00:15:17] with the business.
[00:15:30] Now, coming back to the PDF,
[00:15:34] we've gone over this, what a client actually gives us.
[00:15:37] Prime examples are in the calls folder and the raw intakes.
[00:15:41] Now, one brain, clients inside of it.
[00:15:44] So, this is a really important mental model to adapt to,
[00:15:48] and that is simply that And we do not build a new business brain
[00:15:52] every single client guys, we have one.
[00:15:54] One brain with your company working memory and each client becomes a profile inside
[00:15:59] of it. I'm going to show you how my own business brain looks like.
[00:16:10] And it's really important to understand why does this matter,
[00:16:13] because as a company over Over time,
[00:16:15] we will develop expertise, our playbooks,
[00:16:18] our rules. Our SOPs, our own lessons,
[00:16:21] our hard, our hard-earned lessons that we've had over the,
[00:16:26] our, our own hard-earned lessons that we've had over the past couple of months
[00:16:31] and years of building and working with these different companies.
[00:16:34] We want to save that into the AI and make it live on the
[00:16:38] brain level. So every client that we deal with.
[00:16:41] So every client. So every single client that we've worked with,
[00:16:45] it simply makes the whole system smarter.
[00:16:47] So client five takes you an afternoon,
[00:16:49] not a week.
[00:16:58] And why does that matter? Because our own playbooks,
[00:17:00] our rules, the lessons.
[00:17:03] And why does that matter?
[00:17:05] Our own playbooks, our rules, our SOPs,
[00:17:08] our proprietary knowledge and skills that we develop over time.
[00:17:12] And the hard-learned lessons after having worked with so many clients.
[00:17:16] They all live on the business brain, on the business level.
[00:17:19] And this is what my business brain looks like. So every single client that
[00:17:23] we work with, every single system that we work with,
[00:17:25] they're a bunch of learning lessons, guys. You're going to understand about how not
[00:17:29] to scope a system. How to not use this API,
[00:17:32] what are guardrails that we need to have on the backend,
[00:17:34] fallback methods. Sometimes you'll notice that,
[00:17:36] oh, during a running system, it seemed like OpenAI had an outage time,
[00:17:41] due to which the entire system shut down. Okay,
[00:17:44] so from now on, what we need to do is make the client sign
[00:17:46] up for another API key as a fallback.
[00:17:48] So if OpenAI, for instance, is down, let's say we're using GPT 5.5.
[00:17:54] We need to ensure that in the, that in the rare scenario
[00:17:58] that GPT 5.5 is down, boom,
[00:18:01] there needs to be another model that picks up the work,
[00:18:03] in case it is. And the,
[00:18:06] and the leveraging factor, and the leveraging point over here is that by the
[00:18:10] time you get to client number 5, client number 10,
[00:18:13] it just takes you an afternoon to build the entire system.
[00:18:16] Not in a entire week, because the structure already exists,
[00:18:19] and you're already, and you're just filling a profile.
[00:18:22] Take a look over here, this is what my business brain looks like.
[00:18:25] So what I have is all of my client folders,
[00:18:28] communication, where the AI is trained on how I communicate over WhatsApp,
[00:18:32] my voice notes, email,
[00:18:34] the core knowledge of the business, which is very similar to what the sample
[00:18:39] business brain looks like as well. What is my profile,
[00:18:41] who is my ideal customer profile, what are my personal goals,
[00:18:45] the knowledge, where the knowledge, you can see I build this over time,
[00:18:49] cold email databases, content,
[00:18:51] copy, engineering, image prompts,
[00:18:54] operations, and all of this has been built over years.
[00:18:57] This exact folder over here, it runs my actual company,
[00:19:00] and it has grown, it's a folder that has grown over years.
[00:19:03] Folders, content, playbooks,
[00:19:05] money, and yours will too guys, do not worry about it.
[00:19:08] You don't build these all on day one, you add a folder and new
[00:19:12] files the week you need it, and this happens over time,
[00:19:14] it's a gradual process, as you keep working,
[00:19:17] you keep solving problems and you keep growing the business and getting clients.
[00:19:21] We have every single log, so what this log file is,
[00:19:23] every single activity that has taken place inside of the business.
[00:19:27] my own network, my offer, my platform over here,
[00:19:30] so my Instagram, my LinkedIn, my YouTube,
[00:19:32] my Twitter. My sales playbook,
[00:19:35] and over here we finally have a couple of,
[00:19:38] my, my sales playbook. My sales playbook and the advantages of
[00:19:42] this system is that your AI can suddenly answer across a multiple,
[00:19:47] and the advantage of this system, and the advantage of this system is
[00:19:52] that your AI can now suddenly answer across multiple clients.
[00:19:55] Who is waiting on me? What did I quote them last time?
[00:19:58] Which offer actually landed? And that is when this stops feeling like a folder
[00:20:03] and eventually will start behaving and feeling like an operating system.
[00:20:06] And that is behaving and feeling like an system.
[00:20:23] Now,
[00:20:26] let's begin
[00:20:31] building our
[00:20:37] own brain.
[00:20:42] And if you're thinking that, I do not have a business yet,
[00:20:46] I want you to stop and pause.
[00:20:55] In the previous modules, we picked our starting system in the earlier
[00:21:08] lesson. that choice, the name and a buyer that we can actually hire.
[00:21:12] is a business, written down simply and plainly
[00:21:16] in English. And that is the brain.
[00:21:20] And simply, and simply,
[00:21:22] just having an offer, who we're selling to,
[00:21:25] and what are we building, is exactly how you begin building the brain and
[00:21:29] how the business starts existing. So what we have over here is a simple
[00:21:32] system. Template that I want to be pasting into Cloud,
[00:21:36] which is.
[00:21:57] Now, coming back to the Business Screen
[00:22:02] Template,
[00:22:20] Now, coming back to the Business Screen Template,
[00:22:23] you guys might think. Now, coming back to the
[00:22:28] Business Screen Template,
[00:22:34] it may look like this structure and folder structure that we have over here
[00:22:39] is really simplistic compared to what I have existing right now.
[00:22:43] But, the template is simple enough. On purpose,
[00:22:45] which are essentially just the relevant core files for your business,
[00:22:51] a client folder,
[00:23:33] this folder structure is really simple compared to what I have,
[00:23:36] but the template is really simple on purpose.
[00:23:39] Core files for our business, a client folder,
[00:23:42] and rules for repeated work over here,
[00:23:46] and an output folder as a workbench.
[00:24:11] Where I wanna go over the Claw.md file over here, where if we take
[00:24:14] a look at this file, we are telling exactly the AI on what you
[00:24:16] should be reading. Start from this file, where the Claw.md is the soul of
[00:24:21] the business. That's how you wanna think about it.
[00:24:23] So every single time we open Claw.md over here,
[00:24:26] and we chat something, we ask it a couple of questions,
[00:24:29] the first thing it's going to be doing is read the Claw.md file.
[00:24:32] This is the first file that this AI routes into,
[00:24:35] after which it reads this file and understands,
[00:24:38] okay, I have core over here,
[00:24:40] which explains to me who is the business.
[00:24:42] We have data, which are the numbers,
[00:24:45] Thank Thank you. Thank you. Clients, customer accounts, calls,
[00:24:47] which is the raw evidence, and rules, how repeat jobs get done.
[00:24:51] And output is where the AI is going to be producing different outputs
[00:24:56] for folders and structures. That could be an image,
[00:24:59] it could be a PDF document, it could simply be a markdown file such
[00:25:03] as this, but just information about a client.
[00:25:06] And what we tell the AI guys, really important,
[00:25:09] and take a look at this, we create clear rules,
[00:25:13] because we create clear rules,
[00:25:16] where we're telling the AI, don't read the whole folder for a small question.
[00:25:20] A pricing question needs data pricing dot MD,
[00:25:23] not the entire brain. Because we want to be very intentional with the context
[00:25:28] that we're giving the AI. We do not want the AI to have to
[00:25:31] go and read a hundred different files as we grow this over time.
[00:25:33] For a simple question,
[00:25:35] such as, what is the pricing,
[00:25:37] or what is the expect,
[00:25:40] or what is, for a simple question,
[00:25:42] which could be, what is the exact agreed upon price that I decided on
[00:25:47] the sales call? For questions such as this,
[00:25:49] the AI should instantly know, okay,
[00:25:51] I need to go into calls, find the client name who we've mentioned,
[00:25:55] and open that folder and find the prices. It doesn't
[00:26:00] need to go over each and every single one of these folders and files,
[00:26:03] especially over time as we build this, as we grow this in size.
[00:26:08] And something that's really important are the three rules that we have created over
[00:26:11] here, where rule number one is that one owner per fact.
[00:26:15] So every fact lives in exactly one file where proof exists.
[00:26:19] Pricing, it lives in data slash pricing dot MD because
[00:26:23] the point of this is we want to be very intentional guys about the
[00:26:26] information that we're sending into different files,
[00:26:29] where if you open my business brain, for instance,
[00:26:32] what we have over here is let's just open the folder and the file
[00:26:35] for platform. I have LinkedIn over here,
[00:26:39] right? I have my brandvoice.md where the AI is trained exactly.
[00:26:43] And we give it the information on how am I talking.
[00:27:41] Before, ahem,
[00:27:46] before we go, before we go deeper, and, before we go deeper into,
[00:27:49] And, before we go deeper, into,
[00:27:52] Working with the business brain specifically,
[00:27:55] and instilling, And before we go deeper,
[00:27:58] and before we go deeper into the business brain,
[00:28:01] I really want to visualize how exactly you want to be storing
[00:28:06] context and client information in a really strategic,
[00:28:11] yet easy-to-use,
[00:28:18] and easy-to-use content.
[00:29:00] we go deeper into the business brain,
[00:29:02] I want to go over a really important concept,
[00:29:05] which is context management and file management.
[00:29:08] Because guys, as you saw my folder, over time,
[00:29:11] this exact folder is going to expand significantly in size structure.
[00:29:16] And we need to be really tactical and
[00:29:35] And before we go, and before we go deeper into the business brain,
[00:29:38] I want to go over context management and a really important concept that is
[00:29:42] fundamental to what we're doing over here with AI.
[00:29:45] So let's take a look at this. Let's imagine over here.
[00:29:48] We have our AI LLM, which in this case,
[00:29:51] let's say this is Claude. This is our Claude,
[00:29:54] which is running the entire business.
[00:29:57] If we go back in the terminal.
[00:30:11] Where? What we have is. Is this is a database,
[00:30:14] essentially, of where what we have is this is our entire folder
[00:30:18] structure, which is very adjacent to what we have over here.
[00:30:22] So if you break it up by the different departments and where we store
[00:30:25] information for a company, what we have is we have segregated this and broken
[00:30:30] this into different sections where let's call this section the calls.
[00:30:36] Then what we have is our core information.
[00:30:40] We have our clients over here. Then we have,
[00:30:44] for instance, let's call it our platform, which is our socials.
[00:30:49] We have our knowledge base structure. And this folder structure and files,
[00:30:53] guys, over time, this keeps getting deeper and deeper.
[00:30:57] And what does this most important, where some of you might have even noticed
[00:31:00] in Claw.md, we need to have really specific rules about how exactly
[00:31:05] the AI is routing itself to each one of these specific sections in the
[00:31:09] folder. Let's think about this. Let's say we
[00:31:14] have been in a sales conversation with a hypothetical client and
[00:31:19] we have had call number excuse me,
[00:31:22] we have had call number one with them, we have had call number two
[00:31:26] with them, and we've had call number three, where initially we had the discovery
[00:31:29] session. After that, we got on a deeper scoping call with the stakeholders
[00:31:34] of the company, after which now we're in touch with their procurement.
[00:31:37] After, let's say, the stakeholder meeting went well,
[00:31:41] we have decided, okay, let's move on to next steps.
[00:31:43] And now we are in the process of negotiations.
[00:31:45] Over here. Negotiations.
[00:31:49] And depending on the company size, this could be with their procurement team,
[00:31:53] this could be with their team from a specific department,
[00:31:57] the budget holder, or it could sometimes just be straight away with the CEO
[00:32:01] if it is a smaller company. But regardless,
[00:32:03] let's say in the discovery call, we briefly mentioned,
[00:32:07] because often on calls, clients are obviously,
[00:32:09] or lead leads, they're going to be asking you, okay,
[00:32:11] so could you tell me how much this would cost? And we give a
[00:32:15] rough estimate of somewhere between 7,
[00:32:19] of somewhere between 7 to 12K.
[00:32:22] This is something I do pretty often, guys, especially if this is the initial
[00:32:26] discovery call where we still do not have enough information to really gauge what
[00:32:30] exactly the system is going to be costing. We don't want to slap a
[00:32:33] flat price. And if anything, we always want to anchor higher.
[00:32:36] That is something we go over later on the sales strategies.
[00:32:39] But what I would do is, let's say I say, okay, so typically for
[00:32:41] a system such as this, let's call it a revenue recovery system.
[00:32:44] This would cost somewhere between a 7 to a 12K for a one-time setup
[00:32:48] fee. But depending on scoping it out and understanding
[00:32:53] more of the workflow with your employees and the relevant departments,
[00:32:57] I can give you a finalized price. But now,
[00:33:00] our AI is already going to internally remember probably
[00:33:05] somewhere over here and in the data file,
[00:33:08] which had the pricing information over here.
[00:33:11] It's going to be remembering that, okay, with client,
[00:33:13] let's call them client number A for now. Let's just say this is the
[00:33:16] finalized name of this client. For client number A,
[00:33:20] what we have is we have already decided a price of 7
[00:33:24] to 12K.
[00:33:27] But now, after the stakeholder meeting,
[00:33:31] what we've decided is, cool,
[00:33:33] we, they ask us, can you please send us an official proposal?
[00:33:36] Where what we do is we send them an official proposal
[00:33:41] after getting and collecting all of the information of roughly $9,000.
[00:33:46] And we send this exact proposal over email,
[00:33:52] over here. So now,
[00:33:54] guys, notice this, already there are two sources of information that
[00:33:59] are conflicting for the price. where in one section,
[00:34:02] if the AI is going to be able to be seeing your call transcripts,
[00:34:06] your emails, your slacks, it is going to be thinking that,
[00:34:09] okay, we have a proposal over here of 7 to 12K,
[00:34:14] and we also have something for 9K.
[00:34:16] Now, let's say we have a little bit of back and forth with them
[00:34:19] over email, where they're giving us potentially a little bit of price resistance,
[00:34:23] they're trying to change the price, there are a little bit of,
[00:34:26] a couple of objections coming around, they're saying,
[00:34:28] you know what, could we maybe make it a bit more of a success
[00:34:31] fee, could we reduce the upfront cost,
[00:34:34] this happens, that's normal. And finally,
[00:34:36] after back and forth, we arrive at a final price of 7.5,
[00:34:41] where this is finally what we send the proposal and what we finally agreed
[00:34:45] upon, and this is the invoice that has been issued to them.
[00:34:48] Now, guys, this is really important,
[00:34:51] because this is how it works. Exactly. AI can make mistakes and
[00:34:57] simply lie to your face. This is very common.
[00:35:00] We never want to be trusting and just outsourcing our entire judgement,
[00:35:04] intelligence, and thinking over to the AI,
[00:35:07] because at the end, AI is simply a model that is going
[00:35:11] to be opening the files, understand the information,
[00:35:14] and boom, tell you this is it.
[00:35:16] So, how do we ensure that over time,
[00:35:20] in a very seamless manner, without us having to manually come and tell the
[00:35:24] AI, the AI starts internally thinking that,
[00:35:27] okay, client number A is a client who is going to be between 7
[00:35:30] to 12K, and let's say a week after we had the email where we
[00:35:34] sent the 9K proposal, Hello? The AI now remembers,
[00:35:38] okay, $9,000 is what we're charging client number A,
[00:35:41] and then another one week passes by where we have the back and forth
[00:35:44] with the client, finally arriving at 7.5K and sending the invoice.
[00:35:50] What we need to do over here,
[00:36:25] In order for us to seamlessly
[00:36:29] have the AI update the knowledge and the context of every,
[00:36:33] of every single client. In order for us to have the AI come
[00:36:38] over here and update the context of every single client.
[00:36:41] Not just the pricing, the scope, the system,
[00:36:44] the, the, the communication history we've had with the client.
[00:36:47] What is the expected delivery date? Verbatim on the proposal,
[00:36:50] not what we said initially on the scoping call,
[00:36:53] which is the reason being a scoping call.
[00:36:55] Nothing is finalized. That is us just hypothesizing and assuming that this is how
[00:37:00] we could move forward is we need to have really,
[00:37:02] really strict rules guys, right? Take a look at this.
[00:37:05] What we have is calls and inside of it,
[00:37:07] we have a readme document where let's go over this.
[00:37:10] Only material passed that passed the client data.
[00:37:13] Intake belongs over here. Keep the smallest useful excerpt and de-identify when
[00:37:18] identity is not needed. We have rules over here.
[00:37:21] Raw calls are where facts are discovered,
[00:37:23] not where they live, the moment we discover a piece
[00:37:28] of information.
[00:37:37] We give the
[00:37:43] AI strict
[00:37:48] rules.
[00:37:52] After you drop a transcript here, promote what matters into the file that owns
[00:37:57] it.
[00:38:25] Where
[00:38:33] we need to create the ground structure and the infrastructure for the AI in
[00:38:38] the form of rules to go and update the pricing when the customer changes
[00:38:42] their schedule. That accounts overview.md and history.md.
[00:38:47] The client said, never text after 8pm.
[00:38:49] So Claude should understand this. Say we're trying to automate the customer delivery,
[00:38:53] the customer support of our business. We do not want the client,
[00:38:56] the AI to be messaging or updating the client with project updates by
[00:39:01] itself after 8pm if the client has told us to not do that.
[00:39:04] This is just simple, well-practiced manners in business.
[00:39:09] These are just simple things that allow us to move really cleanly.
[00:39:13] And maintain a really good image with clients and in business overall.
[00:39:17] Because it's really easy guys, to try to automate your business and if in
[00:39:22] anything, get complete slop on the other end.
[00:39:25] Because we know how it feels. Imagine you're working with a client,
[00:39:27] a client who just paid you $5,000, $7,000 and you have AI running your
[00:39:31] company. Where your AI is potentially communicating with the client,
[00:39:35] giving them updates. If there's ever a situation like a clear-cut indication that what
[00:39:39] you've messaged to the client is generated by AI,
[00:39:41] that generally completely deprecates the
[00:39:47] perception they have of yourself. And definitely something you should never do.
[00:39:51] Which is another really important rule. Never automate before testing
[00:39:55] it manually yourself. And having really slow and
[00:40:01] an on-ramp period of the automation. Especially if it's ever client-facing,
[00:40:05] customer-facing, or market-facing. Because your reputation and your
[00:40:09] perception of how others see you totally depends on your every single
[00:40:14] touchpoint you have with them. On messages,
[00:40:16] on a call, over an email.
[00:40:20] So we will get over having the AI manage your communication with clients,
[00:40:24] your content. But it is really important to always keep in mind,
[00:40:27] and the only way we can bypass that,
[00:40:30] from the AI, because once it has access, guys,
[00:40:32] to our Slack channel, our WhatsApp, our email,
[00:40:34] who's stopping the AI to go ahead and do all of that?
[00:40:38] And just reach out to clients and message them. And that is why we
[00:40:40] have rules over here. Now we have another really interesting,
[00:40:44] simple sentence. When we're off-boarding a client,
[00:40:47] follow the client agreement and return the return or delete calls as required,
[00:40:51] remove the working copies, revoke access, and confirm completion to the client.
[00:40:55] Which is very normal.
[00:41:37] creating systems out of it. One thing I do want you guys,
[00:41:40] and I heavily encourage you to do, is go over each and every single
[00:41:44] one of these files. The more you familiarize yourself with
[00:41:49] the structure of a .md file. This is,
[00:41:52] guys, this is how AI communicates to itself.
[00:41:55] So the more you can familiarize yourself with it,
[00:41:58] the rules.md files, the output,
[00:42:01] and just these readme files, the better you're going to be incorporating
[00:42:06] and understanding how to use this business brain. And the much faster you're going
[00:42:09] to be able to make progress with it, not just in terms of creating
[00:42:12] systems, but creating valuable systems for clients that get really good results.
[00:42:17] Because this is how AI architecture really is formed underneath the hood.
[00:42:21] It is .md files, the .pricing file,
[00:42:24] and you can see most of these files, guys, are going to be written
[00:42:26] completely by AI. This entire file has been written by AI.
[00:42:29] In my business brain, every single thing is created completely using AI.
[00:42:34] All these files over here. My job is to simply be the architect of
[00:42:38] this system over here. The more you understand from first principles how this
[00:42:43] works, and the more you actually use these systems,
[00:42:46] you will be able to front run and see ahead in time.
[00:42:50] The common mistakes that AI tends to have,
[00:42:53] and that will allow you to create systems that are much more accurate,
[00:42:57] faster, and allow you as a business to move much faster.
[00:43:00] So, let's get into the Google document over here,
[00:43:04] where what we have is creating our own brain.
[00:43:06] So, if you're thinking that I don't have a business yet,
[00:43:09] I want you guys to stop and think back to the previous lessons we
[00:43:13] went over. Where we picked a starting system in the earlier lesson.
[00:43:18] That choice, the exact system, whether it was an ad creation system,
[00:43:22] a revenue recovery system, an owner's dashboard,
[00:43:24] the name, and a buyer, we actually read,
[00:43:28] or reach. That is the business, guys.
[00:43:30] A business is simply having an offer,
[00:43:32] selling it, and making money, cash flow.
[00:43:34] That is a business in very simple terms. Now,
[00:43:37] the bridge is how the business starts existing and evolving and scaling over time.
[00:43:42] So what we have over here is we have a really simple prompt,
[00:43:45] and now we're finally going to get into using AI along with the system.
[00:43:48] Well, let's just give it a quick read, and then I'm going to copy
[00:43:50] this and show you how it works. So read the blank template and its
[00:43:54] instructions inside business brain template.
[00:43:57] So we tell the AI that this is the folder you should be referring
[00:43:59] to. Copy the entire business brain template template folder to a new folder named
[00:44:04] my brain. Keep it and fill my brain score files for my business.
[00:44:08] From what I tell you here, nothing else. Do not invent achievements,
[00:44:11] clients or numbers. I do not state guys.
[00:44:15] Another thing I really want to go over is do not emphasize too much
[00:44:18] about prompting at this age and model.
[00:44:21] Yes, prompting was something that was very relevant about two years ago.
[00:44:25] However, Good. In this age and era,
[00:44:28] the LLMs are more than capable of being able to prompt and create prompts
[00:44:33] for themselves. You don't know. I do not want you guys to be going
[00:44:36] deep into different prompt playbooks and trying to become a prompt engineer.
[00:44:40] That is really a low leverage skill in this day and age.
[00:44:43] And for what we're trying to achieve, the AI is simply the vehicle for
[00:44:47] us to grow. Scale companies make money.
[00:44:50] So, what we put is,
[00:44:52] who am I? We are going to fill this out, so let's open this
[00:44:54] over here, and let's go over this. So,
[00:44:57] let's give this a close. Now, finally, we open the terminal,
[00:45:00] and we click here, and we paste this.
[00:45:03] So, as we can see, very often with Claude,
[00:45:06] what we're going to do is, when you have a really big file of
[00:45:09] 200, 300 plus characters,
[00:45:13] it's going to truncate it like this. So,
[00:45:15] what we want to do is, if you're trying to paste something directly into
[00:45:18] it, what you can do is, open it up.
[00:45:26] Cool, cool. So now that we have this established,
[00:45:29] let's go on to making the brain. So,
[00:45:32] if you're thinking, number one, I do not have a business yet.
[00:45:35] I want you to mentally pause and think back to the previous lessons that
[00:45:39] we went over, where we picked our business.
[00:45:42] Our starting system in the earlier lessons,
[00:45:44] whether that is an ad creative system,
[00:45:48] a owner reporting and dashboard led system,
[00:45:51] revenue recovery, that choice,
[00:45:54] guys, the name and the buyer that we've decided is who
[00:45:59] we can actually reach, that is the business.
[00:46:02] A business, simply put, guys, is we have an offer,
[00:46:05] we sell that. and we make cash flow and revenue.
[00:46:08] That is, simply put, what a business is.
[00:46:11] So the brain is how the business starts existing and evolving over time.
[00:46:16] So here's the first exercise. What I have over here is a prompt that
[00:46:21] I want us to paste and for everyone watching this video to copy into
[00:46:25] your cloud. Quick disclaimer,
[00:46:28] we're not going to be using templated prompts such as.
[00:46:43] So let me do what I'm asking.
[00:46:46] So let me do. So first, let me do what I'm asking you to
[00:46:47] do out of the menu from that lesson. Let's say that the offer that
[00:46:51] I picked is I do owner reports and revenue reports.
[00:46:54] Revenue follow ups for local businesses. That is my starting point.
[00:46:58] I finally decided and that decision is going to be our scenario for the
[00:47:02] rest of this course. Every lesson from here on builds on that exact business.
[00:47:08] So what we have over here is I have this exact prompt where we
[00:47:10] simply copy it. And what I've done over here is I've pasted it.
[00:47:14] It's the exact same prompt. But what I've done is I've filled in the
[00:47:17] blanks where I want you guys be putting in your name,
[00:47:19] what you do, who are you selling and what are you selling,
[00:47:23] to whom is it, and what is your goal. Simply put,
[00:47:26] I said I'm Aryan, I run an AI company for scaling businesses.
[00:47:30] What I'm selling is owner reports and revenue follow ups for local businesses,
[00:47:34] local clinics and service business owners in my city,
[00:47:36] Los Angeles. And my goal for the next 90 days is to get one
[00:47:40] paying client. So what we want to be doing here over here is copying
[00:47:44] this and pasting it.
[00:47:53] Where we open Claude and simply put we just paste it into the simple.
[00:47:58] Now let's take a look at what Claude gives us over here.
[00:48:00] You can even notice guys, the exact files.
[00:48:03] Take a look at this. The first thing it read like I told you
[00:48:05] was the Claude.md file. Which I mentioned is the soul of the AI
[00:48:10] and that is why it is so important to have this Claude.md file.
[00:48:15] Accurate, intelligent with the right structure which we're going to be creating and evolving
[00:48:20] over time. But for now, let's just take a look at what the AI
[00:48:23] does over here.
[00:48:30] So the
[00:48:35] AI finished running and let's take a look at it.
[00:48:37] So it read the whole template, the Claude.md,
[00:48:40] and the core, four core files, the data,
[00:48:42] clients, calls, rules, and outputs.
[00:48:45] And it copied the intact structure of the template into my brain over here.
[00:48:51] Verified the template is byte-identical and we can take a look.
[00:48:54] Claude.md. Let's open these folders guys,
[00:48:56] Claude.md, convert it to preview.
[00:48:59] And we can see for my business name,
[00:49:01] you can fill it up, runs by Arin that helps businesses scale.
[00:49:04] The current focus is over here and we can tell it the
[00:49:09] core, as we can see,
[00:49:11] it is routing it to the relevant folder. And
[00:49:16] we can check these files, we have business.md,
[00:49:18] where we can see Arin runs a one-person AI company,
[00:49:21] two things both aimed at local services,
[00:49:23] and here is the information. And that is it guys.
[00:49:27] We have customers.md, we have goals.md,
[00:49:29] and over time, we enhance, improve,
[00:49:32] and deepen the structure of these folders and files.
[00:49:35] But that is the company on paper. Under a minute.
[00:49:38] Nobody is waiting for you to feel ready and the files don't care.
[00:49:41] And notice guys what it didn't do. It didn't invent client details I don't
[00:49:45] have, and it left the open questions,
[00:49:48] completely open.
[00:50:02] This can be deepened over time through sales conversations
[00:50:07] we have. We push it through the AI and the conversations we couldn't
[00:50:11] close. We will understand better who is not the type of customer we are
[00:50:14] looking for. What are patterns and trends we can see in people and potential
[00:50:18] leads in the future. So we don't waste time on them.
[00:50:21] Because there is going to be a specific segment and a type of a
[00:50:24] persona, company and people.
[00:50:26] Who are not suited to be our customers for our specific offer.
[00:50:30] Either they are tire kickers, they waste time,
[00:50:32] they're indecisive or simply put,
[00:50:35] they're just looking around for different tools but not really ready to make a
[00:50:39] significant buying decision. Now,
[00:50:42] coming back to the file over here, what we want to be doing is
[00:50:45] now incorporating the client into this.
[00:50:48] Remember the client? The MedSpa Company.
[00:50:50] What we're going to be doing is pasting this exact prompt into the AI.
[00:51:14] So we simply copy this,
[00:51:18] and we come over here, and paste it.
[00:51:20] And let's give it a run. So,
[00:51:22] while this is giving a run, it's gonna take a minute or so.
[00:51:26] We can see, that, when I'm myself.
[00:51:30] So while we're giving this a run. Guys,
[00:51:33] so while we're giving this a run. So while we're giving this a run.
[00:51:37] I wanna preface that, in my own day to day.
[00:51:40] I do not paste prepared prompts. I just talk to the AI.
[00:51:43] I use a tool over here, called WhisperFlow.
[00:51:46] That is a simple dictation tool. Where all we do is,
[00:51:49] I click on a hotkey on my computer. And as we can see,
[00:51:52] it's listening to me speak. And I tap it, and boom.
[00:51:55] There we go. It is an excellent tool for productivity.
[00:51:58] And getting things up and running really quick.
[00:52:01] This prompt over here that we have, this is just training wheels.
[00:52:04] It exists so that your first build is scoped.
[00:52:07] It is safe and checkable. Later in the course,
[00:52:10] you'll be yapping at it, like I do pretty much.
[00:52:12] And once you get really comfortable with the architecture and the structure of this
[00:52:16] folder and the business brain. One more thing I want you guys to notice,
[00:52:20] is take a look at this. the client's folder over here.
[00:52:23] It lives completely in this
[00:52:49] One thing I want you to notice over here is that the client's raw
[00:52:53] files, it lives outside of the brain.
[00:52:56] This is a separate folder that we have over here and the client folder
[00:53:00] is completely separate. It's isolated and that is very normal.
[00:53:05] What the client sent to you is a pile. It can sit anywhere.
[00:53:08] What is important is that the AI,
[00:53:11] in this case Claude, has access to it.
[00:53:13] And it can pull information, distill information,
[00:53:16] and route it back into the brain. The cited truth,
[00:53:21] the exact verbatim information,
[00:53:23] and that distinction is what keeps our business brain over here trustworthy forever.
[00:54:20] And while the AI is, and while the AI is still working on the
[00:54:21] prompt, what I want to take a look over here are a couple of
[00:54:25] really special features. specific lines that we have very intentionally fed into the AI.
[00:54:29] Take a look at this. Do not read company one medspa brain or company
[00:54:33] one data. That completed client folder is the comparison for later.
[00:54:37] So number one, the reason for existence of these two files over here is
[00:54:41] for the next video, the next exercise that we do.
[00:54:44] But what we're doing over here is something that should be a mental discipline.
[00:54:49] That we always walk through when we're using the AI,
[00:54:52] we need to guide the AI, like I said before,
[00:54:55] about which files it may read and which ones it not.
[00:54:58] We are the, we are the ones who architect and run the system.
[00:55:02] So we are the ones who decide and tell the AI what it is
[00:55:04] allowed to do and what not to do. We control the evidence.
[00:55:08] Really, really important. We never once again want the AI to just go,
[00:55:22] Ballistic. We never want the AI to go fully ballistic, just running over everything
[00:55:24] and doing as it see fit. Then finally,
[00:55:27] what we have is, take a look at this, in pricing.md,
[00:55:29] every price statement with its date and exact source line.
[00:55:33] The reason being, remember the previous example I spoke about,
[00:55:37] how we, We want to be distilling the information from a conversation
[00:55:41] we've been having over time with a client, where we might have sent them
[00:55:44] an initial scoping proposal, then we send the official proposal,
[00:55:48] and after back, back and forth negotiation,
[00:55:50] we arrive at a final price. And that is why,
[00:55:54] we want the system to record when someone said something on a specific date,
[00:55:58] not a permanent approved fact. This is really important.
[00:56:02] What this single sentence allows us to do,
[00:56:05] is this is the difference between a system that documents evidence,
[00:56:08] and one that understands when it heard the last,
[00:56:13] in brackets, truth. By simply having a structure
[00:56:18] in the files that, okay, on the 10th of September,
[00:56:21] we quoted them between 7 to 12K,
[00:56:23] then on the 20th of September, we sent a proposal for $9000,
[00:56:27] and finally, on the 30th of September is when we send the official invoice
[00:56:31] for the 7.5K. That is a much better way of ingesting
[00:56:36] and keeping your AI systems up to date with all of the information.
[00:56:41] As we can see, as we can see,
[00:56:43] the AI is ingesting about 80 pages of transcripts.
[00:56:47] So I'm going to cut ahead to the video,
[00:56:50] but in a real engagement, this is the part when you typically go grab
[00:56:54] coffee or a snack while the AI is working on every single thing.
[00:56:58] After 8 minutes of hard work, the AI has finally completed it,
[00:57:02] and let's take a look. Over here, we open MyBrain and Clients.
[00:57:06] We finally have every single thing. for client number one.
[00:57:09] Citrine Val instilled over here,
[00:57:11] Overview, Pricing, Questions for Owner,
[00:57:14] Rules, and Source Check. And we can even take a look at the output
[00:57:17] that Claude, or Codex in this instance,
[00:57:20] will always give us in the terminal. And let's take a look at something
[00:57:23] really interesting guys. What it did with the pricing,
[00:57:26] because this is the most important moment of the video.
[00:57:29] Disagreements, I kept visible. $13 vs $12 in the July
[00:57:34] 1 call says the menu said $13.
[00:57:37] A $12 sheet was under the card reader the day before that.
[00:57:41] Both recorded, official answer left as Marisol confirming against the menu which
[00:57:45] is what she asked for twice. And what we can confirm this with is
[00:57:49] open over here, Citrine Val, and open the pricing.md file.
[00:57:53] This is exactly what the AI has created for us.
[00:57:56] And we can see there is a discrepancy because the owner themselves
[00:58:01] said $13 a unit on the July 1st call.
[00:58:07] There's also an old $12 sheet still physically floating around her front desk.
[00:58:11] So the two different sources of information,
[00:58:14] really important, conflicting pieces of information.
[00:58:17] And the profile over here recorded both of them with the dates and the
[00:58:22] sources. And that leaves the final call with the owner.
[00:58:25] It did not pick one. And this is exactly how we prevent the AI
[00:58:29] from making up information. The AI says it itself.
[00:58:32] It is a record of who said what number, on what date,
[00:58:35] and in what conversation. None of these numbers is an approved price because the
[00:58:39] document that holds the approved price, the clinic's current menu,
[00:58:42] is not among the sources. And it,
[00:58:45] tells us over here, Marisol,
[00:58:47] she set the rule twice herself. One from a memory,
[00:58:51] and one from the actual call conversation,
[00:58:53] which is $13, and the other one was $12 a unit.
[00:59:09] And the impressive part is not the AI catching this.
[00:59:15] In today's world,
[00:59:18] a good model would probably catch it on its own.
[00:59:21] That is not the point. The point is that the brain and the structure
[00:59:24] of the system, it makes it guaranteed instead of probable.
[00:59:28] So any model, any day, any phrasing,
[00:59:31] any source of information, the system stops here.
[00:59:34] So now what we're not selling is no longer just intelligence.
[00:59:38] What we're selling is certainty coupled with intelligence.
[00:59:41] And now I want to take a look at this file over here as
[00:59:43] well. Questions for the owner. These are still calls that we have to clarify
[00:59:47] with them before we can close the deal. Because we cannot start building a
[00:59:51] system and invoice the client until we do not have the exact details verified.
[00:59:56] Because we do not want to invoice the client guys.
[00:59:59] Take the money and start building something.
[01:00:02] And later on in the project, understand these little details that might completely change
[01:00:06] the architecture of the system. A simple fact or detail about a CRM they're
[01:00:10] using, about the way they use a specific tool.
[01:00:13] It might change the complete scope of the system we're building,
[01:00:15] where we expected to have a system built within five business days.
[01:00:19] But now, due to this one little detail that we realized two days into
[01:00:22] delivery, we're like, oh shit, all of a sudden this is going to take
[01:00:26] me 20 extra days to build the entire system.
[01:00:35] And this file over here, this might just be my favorite,
[01:00:39] because a good profile knows what it doesn't know.
[01:00:42] And every one of these questions is very highly likely to be a question
[01:00:46] that would have been a silent wrong guess in most people's AI setup.
[01:01:09] Taking
[01:01:14] a look once
[01:01:21] again at
[01:01:27] the prompt we pasted before, what the client profile must contain. The wording differs
[01:01:30] between runs, but the substance, the verbatim facts,
[01:01:33] they must be not. We must incorporate and digest all of this information.
[01:01:37] The Pricing.md file. It records the Botox as $13 a unit,
[01:01:41] as the owner's statement on the July 1st call,
[01:01:44] but notes that the old $12 sheet still circulates at the front desk and
[01:01:48] it leaves the official current menu in limbo as the owner's
[01:01:52] confirmation. A profile that silently picks one is completely wrong guys.
[01:01:56] That means that you're building your API. Based on the wrong data,
[01:01:59] which is going to obviously produce the wrong output and not deliver upon the
[01:02:03] expected promise. Then we have the Rules.md file,
[01:02:06] the questions for the owner that holds the real unknowns,
[01:02:09] the current official menu and the Google connector.
[01:02:11] The Source.check.md and the business brain template.
[01:02:14] So now what I want to do is I want to go over a
[01:02:16] couple of more prompts as a quick exercise about us using the system.
[01:02:20] The person who is not there. So, if we go over the call transcripts,
[01:02:24] there is a person, a call mentioned a name with no
[01:02:28] story behind it. And this is common once again.
[01:02:31] Guys, sometimes you're asking your clients if you're building CRM,
[01:02:35] sales activation systems, you're going to be asking them for their sales calls.
[01:02:39] And as you could expect, their sales calls could come in any shape or
[01:02:43] form. It could be incredibly unstructured or you might be really lucky.
[01:02:46] And what they might provide to you is a Firefly.ai API
[01:02:51] link, which is very cleanly recorded with the names of every single
[01:02:56] person on the meeting with the AI transcripts,
[01:02:58] all accessible through an API. But this is wishful thinking because a lot of
[01:03:03] times, especially when you're working with old school fashioned,
[01:03:06] old school service based businesses, they do not have such systems,
[01:03:10] put simply. So let's copy this. We're called to mention someone called Renata B.
[01:03:14] For this question only, you may also search company one medspa.
[01:03:17] So we're telling it, you can search the CRM and tell me what the
[01:03:20] evidence actually supports about Renata. And this is common guys.
[01:03:23] Like a lot of times when I am going deep into a company where
[01:03:27] they've provided me all of the information and I'm still scoping it.
[01:03:30] I'm trying to understand what am I building over here?
[01:03:32] What am I doing? And this A lot of times I will just come
[01:03:34] into the AI and I just talk to it. I'll be using this workflow
[01:03:38] and I just tell it like, okay, we have all of this information.
[01:03:41] Can you just break everything down for me? Like I'm confused.
[01:03:43] I just tell it very simply put where if I have,
[01:03:46] let's say company one medspa and I have a lot of different files,
[01:03:49] I'm going over them quickly.
[01:03:52] I'm scrolling through it. I just turned on my mic and I'm just asking
[01:03:54] it, okay, who is this person? What did they mention over here?
[01:03:57] Sometimes I'm on a sales call. I'm making notes myself on a pencil,
[01:04:01] let's say, or on my notepad about things that are a little like just
[01:04:05] extra talking points that I want to clarify myself with the AI and that
[01:04:09] is exactly how I'm going to be doing it. So over here,
[01:04:12] something that I'll be doing is I'll be switching the model to a faster
[01:04:16] model because I want to be moving a little bit faster and this is
[01:04:18] how you want to be choosing which LLM models to use.
[01:04:21] For simple tasks such as this guys, we do not want to be using
[01:04:24] Fable or Opus. Simply said, it's just going to take too long and I'll
[01:04:27] even draw down the effort to medium because all we're doing is a very
[01:04:31] simple search query task. Let's press enter and we're just telling it like all
[01:04:34] you have to do is go into the CSV file over here,
[01:04:37] extract this information and break it down for me.
[01:04:40] What is Renata B? Who is this person?
[01:04:42] Is it a person? Is it a product? Is it a company?
[01:04:45] And that really allows us to move with speed while ensuring
[01:04:51] that we have full understanding of the system we're building.
[01:04:54] And as we can see over here guys, the system responded really quickly simply
[01:04:58] by changing the model to Sonnet 5 and reducing the effort of thinking.
[01:05:02] It tells us, I searched for Renata's case insensitive and
[01:05:07] what it actually supports. It's over here, what it supports that Jordan asking Marisol
[01:05:11] to search the 12th, mass training sample for Renata because a name Renata B
[01:05:15] appeared on the line. Okay. And I quickly typically just go through this to
[01:05:20] a call transcript and what it does not support.
[01:05:24] Is that Renata B is a real current or former patient.
[01:05:27] Could be a patient. She could be a staff reference and old note or
[01:05:30] something else entirely. And this is exactly what I said before we put in
[01:05:34] the prompt where I, myself, I'm like,
[01:05:36] okay, sometimes on sales calls,
[01:05:39] especially when you're working with industries that are newer.
[01:05:42] And this is something I've experienced myself, guys. I've worked over the past two
[01:05:45] years. I've worked with hedge funds. I've worked with consulting firms.
[01:05:48] I've worked with dentistry,
[01:05:50] med spas. I've worked with soccer clubs. Academies,
[01:05:52] NBA clubs. I've been like,
[01:05:54] so all over the place. And naturally when you're entering different industries,
[01:05:57] they have internal lingo and wording that these guys use themselves.
[01:06:02] And it is our job to understand this information on a high level ourselves
[01:06:06] and not once again, automate all of the thinking to the AI,
[01:06:10] because we still need to understand these simple concepts.
[01:06:13] And this is a prime example.
[01:06:15] Where simple concepts,
[01:06:17] words may come up on a call. And in this case,
[01:06:20] it is something called a Renata Bee. And we just tell the AI,
[01:06:24] and we just tell the AI, what is Renata Bee?
[01:06:27] And as we can see over here, we got a dead end answer from
[01:06:31] the AI saying that after going to 2400 CRM
[01:06:36] records, the answer is, I honestly don't do not know.
[01:06:39] And here's exactly what I will not invent.
[01:06:41] I love that over here. This is beautiful.
[01:06:44] The refusal is the product, guys. We are creating business truth and business systems.
[01:06:48] This cannot be something that is artificially created or the AI is just free
[01:06:53] running and doing whatever it feels like. And this system fills that exact gap
[01:06:57] with plausible guesses in the system and text the stranger about someone
[01:07:02] else's business. It's medical treatment because if this is an outbound system where this
[01:07:06] is reaching out to different clients or it is a revenue recovery system,
[01:07:10] which is the offer that I picked for myself. The last thing I want
[01:07:13] the AI is to start sourcing a made up word,
[01:07:17] a product, a client, which we still do not know as Renata B and
[01:07:21] putting that information in the email copy that we're sending out the LinkedIn posts
[01:07:25] or the LinkedIn emails. So when you demo it to a real owner,
[01:07:29] this moment of the system refusing is what makes them trust you and build
[01:07:33] instant credibility because now they know that I can give,
[01:07:37] I can give the keys to my business to someone.
[01:07:39] And I know for a fact that they're not going to mess it up,
[01:07:42] create something customer facing that tarnishes our reputation.
[01:07:46] Coming back to the document.
[01:07:49] Now,
[01:07:52] coming back
[01:07:59] to the document,
[01:08:08] this exact exercise is super simple, but this is, once again, a mental framework.
[01:08:11] This is how we want to be thinking when we're working with AI.
[01:08:14] And we can see, the correct answer is one dead end line in Call
[01:08:18] 2. Zero matches across the CRM,
[01:08:21] and the agent almost knows nothing.
[01:08:23] So, instead of inventing a story, zero matches does not prove that she never
[01:08:28] existed.
[01:08:37] So,
[01:08:46] coming back to the document, this was a really simple exercise,
[01:08:50] guys. But, this is more of a mental framework that we always
[01:08:55] want to be operating by, where we never let the agent invent
[01:09:00] a history, fabricate data, details,
[01:09:02] and we do not accept it. And once again,
[01:09:05] we are not going to be the ones who are going to be reviewing
[01:09:08] every single piece of information here. The magic is we instill the frameworks
[01:09:12] into the AI, the thinking patterns, the rules,
[01:09:15] like we already did. And that is how the AI checks itself and guardrates
[01:09:20] itself, by QA-checking its own data,
[01:09:23] to verify that we are not making up details.
[01:09:25] And this is exactly how we ensure that a real system never
[01:09:30] texts a stranger or never texts a client or a prospect with made-up information.
[01:09:35] I'll like you guys, when you're going over this exercise,
[01:09:38] check a citation yourself. Once you run the initial prompts,
[01:09:41] you will go into MyBrain, the clients, and you will see there's a file
[01:09:44] called SourceCheck.md, which we also have one created over here.
[01:09:48] Pick the Botox line and go over this. Now,
[01:09:50] finally, what I want us to do is let's quickly compare this with the
[01:09:54] completed profile. So, let's go back to opening.
[01:10:22] And, finally, I want us to.
[01:10:25] And, finally, I want you guys. And, finally, I want you to compare this
[01:10:26] again. And, finally, I want you to compare this against the completed profile.
[01:10:31] If we come over to our initial database,
[01:10:33] what we have is Company 1 MedSpa and you can see we already have
[01:10:36] a brain over here.
[01:11:01] And
[01:11:07] finally, to close off this lesson, what I want you to do is go
[01:11:11] back into the database and open Company1 MedSpa.
[01:11:14] You will see that we already have a completed version of this profile in
[01:11:18] your downloads. And what I want you to do is genuinely open this
[01:11:22] up and compare it against yours.
[01:11:25] There will be different versions. That is normal.
[01:11:27] The different wording is fine. But what you also notice is
[01:11:32] that both of them preserve the price disagreement instead of laundering it and making
[01:11:36] it up.
[01:11:43] And that is really the core fundamental of this lesson.
[01:11:47] This completed profile that we have created and the one you will be provided
[01:11:50] with is also what the next lesson is going to cover.
[01:11:54] What we're going to be doing is create a owner facing report as a
[01:11:57] deliverable and what a potential system that we will be creating in this case
[01:12:02] would look like. So finally, this is where we are.
[01:12:05] We have a business brain. We have a client inside of it and a
[01:12:09] profile that admits what it doesn't know. So the next lesson now,
[01:12:13] this exact folder is starts learning. It produces a report that the business actually
[01:12:17] pays for. And this is how we start making progress.
[01:12:21] Now, finally, one thing that I want you to do is we
[01:12:26] have created a 15 minute task in the practice and help tab,
[01:12:30] which is over here. I want you to go over this.
[01:12:33] And this is a personal exercise that will allow you to
[01:12:37] sharpen and strengthen your skills. So run through this entire task over here,
[01:12:42] find where the build overreached, and there's always one claim that goes past the
[01:12:47] source. Finding that exact touch point,
[01:12:49] that exact thing where the AI is claiming something that is not true.
[01:12:53] It is a skill and something that that is very valuable
[01:12:58] to allowing us to create a business brain and scaling this model.