WEBVTT 00:00:00.000 --> 00:00:03.700 Here is a pretty brutal truth about the AI agency 00:00:03.700 --> 00:00:09.259 world right now. Most of them are slowly starving 00:00:09.259 --> 00:00:11.359 to death. They're just surviving on this diet 00:00:11.359 --> 00:00:15.019 of $1 ,500 projects. Exactly. I mean, they build 00:00:15.019 --> 00:00:17.339 a cool little one -off chat bot, the client plays 00:00:17.339 --> 00:00:19.980 with it for a bit, but then they just get bored, 00:00:20.079 --> 00:00:22.600 they drop it two months later, and that whole 00:00:22.600 --> 00:00:25.489 exhausting cycle just repeats itself. Welcome 00:00:25.489 --> 00:00:27.530 to our deep dive. I am really glad you're joining 00:00:27.530 --> 00:00:29.789 us today. If you are listening to this and feeling 00:00:29.789 --> 00:00:33.750 that exact agency burnout Just take a breath. 00:00:33.950 --> 00:00:36.750 Yeah, we've been there. We are looking at a massive 00:00:36.750 --> 00:00:38.789 fundamental shift today We're moving completely 00:00:38.789 --> 00:00:41.810 away from selling those small disconnected automation 00:00:41.810 --> 00:00:43.890 right because you have to we are talking about 00:00:43.890 --> 00:00:46.520 building an actual intelligence foundation An 00:00:46.520 --> 00:00:49.859 AI operating system or an AIOS. Which transforms 00:00:49.859 --> 00:00:53.000 those exhausting one -off projects into serious 00:00:53.000 --> 00:00:55.299 foundational retainers. We're talking $2 ,500 00:00:55.299 --> 00:00:58.619 to $6 ,000 a month. That's a huge jump. It is. 00:00:58.840 --> 00:01:01.020 So today, we're going to explore exactly why 00:01:01.020 --> 00:01:03.359 that old model fails. We'll unpack the three 00:01:03.359 --> 00:01:07.019 specific layers of this AIOS. We'll look at how 00:01:07.019 --> 00:01:10.180 to actually wire it to real, grieving business 00:01:10.180 --> 00:01:12.859 data. And finally, we'll break down the exact 00:01:12.859 --> 00:01:15.340 pricing and contracts you need to protect this 00:01:15.340 --> 00:01:17.989 model. It really is a completely different ballgame. 00:01:18.109 --> 00:01:20.069 But before we can build the solution, we have 00:01:20.069 --> 00:01:22.230 to understand why the current agency model is 00:01:22.230 --> 00:01:25.230 just breaking down, specifically for small and 00:01:25.230 --> 00:01:28.150 medium businesses. Well, the old model just has 00:01:28.150 --> 00:01:32.159 this brutal hard ceiling. I mean... Every single 00:01:32.159 --> 00:01:35.340 new automation project requires its own custom 00:01:35.340 --> 00:01:38.260 setup. It needs its own discovery phase, its 00:01:38.260 --> 00:01:41.819 own ongoing maintenance. It's so manual. So as 00:01:41.819 --> 00:01:44.420 your client list grows, your profit margins actually 00:01:44.420 --> 00:01:46.379 shrink. You're basically just doing more manual 00:01:46.379 --> 00:01:48.620 labor to keep the bots running. Exactly. But 00:01:48.620 --> 00:01:51.840 here is the really fascinating part. These small 00:01:51.840 --> 00:01:54.519 and medium businesses are incredibly ready to 00:01:54.519 --> 00:01:57.620 pay. They absolutely are. A 2026 report on AI 00:01:57.620 --> 00:02:00.430 spending shows very clear numbers. They're currently 00:02:00.430 --> 00:02:04.290 spending $500 to $5 ,000 a month on ready -made 00:02:04.290 --> 00:02:07.549 AI tools. Wow. Or they're shelling out $30 ,000 00:02:07.549 --> 00:02:10.310 to $100 ,000 on these massive custom builds. 00:02:10.590 --> 00:02:12.250 So the money is absolutely there. I mean, the 00:02:12.250 --> 00:02:14.569 budget exists. The real question is who they 00:02:14.569 --> 00:02:16.830 actually hand that check to. Right. Because the 00:02:16.830 --> 00:02:20.129 current scattered solutions out there are causing 00:02:20.129 --> 00:02:23.189 two massive pain points for these founders. Yeah. 00:02:23.270 --> 00:02:26.449 Pain number one is context loss. Yes. I want 00:02:26.449 --> 00:02:28.710 to dig into that. Think about a typical founder. 00:02:28.849 --> 00:02:32.069 They sit down to work. and they end up wasting 00:02:32.069 --> 00:02:35.050 like 10 minutes every single session, just re 00:02:35.050 --> 00:02:38.169 -explaining their business to ChatGPT or Claude. 00:02:38.250 --> 00:02:40.250 Every single time. Right. They have to retype 00:02:40.250 --> 00:02:42.789 their target audience, they have to explain their 00:02:42.789 --> 00:02:45.370 team structure, their specific products. Only 00:02:45.370 --> 00:02:49.150 then can the actual real work start. It's kind 00:02:49.150 --> 00:02:51.590 of like stacking Lego blocks of data every single 00:02:51.590 --> 00:02:53.629 day. That's a great way to put it. But every 00:02:53.629 --> 00:02:55.659 morning, someone comes along and just... kicks 00:02:55.659 --> 00:02:57.860 them completely over. You have to start from 00:02:57.860 --> 00:03:00.740 scratch. That is a perfectly vivid way to picture 00:03:00.740 --> 00:03:02.699 it, and that leads right into pain number two, 00:03:02.780 --> 00:03:05.400 which is this massive tool sprawl. A typical 00:03:05.400 --> 00:03:10.120 small business pays for so many separate, completely 00:03:10.120 --> 00:03:13.699 isolated tools. We've got a CRM over here, an 00:03:13.699 --> 00:03:16.039 email marketing tool over there. Right, and an 00:03:16.039 --> 00:03:18.580 ad platform, an analytics dashboard, plus two 00:03:18.580 --> 00:03:20.479 or three different AI subscriptions. None of 00:03:20.479 --> 00:03:22.669 these tools. talk to each other automatically. 00:03:22.889 --> 00:03:26.069 No, they don't. So the human becomes the API. 00:03:26.490 --> 00:03:28.789 The human glue. Exactly. The founder is just 00:03:28.789 --> 00:03:31.169 manually copying and pasting data between tabs 00:03:31.169 --> 00:03:34.169 all day. In fact, 40 % of these businesses name 00:03:34.169 --> 00:03:37.409 integration as their absolute biggest roadblock 00:03:37.409 --> 00:03:40.949 to using AI effectively. So does moving to an 00:03:40.949 --> 00:03:43.830 operating system model actually eliminate that 00:03:43.830 --> 00:03:46.169 daily context loss, or does it just hide it? 00:03:46.389 --> 00:03:49.770 It eliminates it by giving the AI, a permanent 00:03:49.770 --> 00:03:51.870 readable memory bank. OK, so we know the core 00:03:51.870 --> 00:03:54.530 pain is a total lack of permit memory. Let's 00:03:54.530 --> 00:03:56.590 look at the exact architecture needed to fix 00:03:56.590 --> 00:03:59.349 this. Yeah, so the AIOS architecture has three 00:03:59.349 --> 00:04:02.229 distinct layers. Layer one is the context OS. 00:04:02.449 --> 00:04:05.490 That is your base. Layer two is the integrations. 00:04:05.590 --> 00:04:07.590 Think of that as the plumbing. And layer three 00:04:07.590 --> 00:04:09.810 is the workspace. Let's briefly touch on that 00:04:09.810 --> 00:04:12.069 workspace layer first. That is basically the 00:04:12.069 --> 00:04:14.349 interface. Yeah, mostly. For the builder, like 00:04:14.349 --> 00:04:16.750 the agency owner, that usually means using a 00:04:16.750 --> 00:04:18.759 coding environment. something like Claude Code 00:04:18.759 --> 00:04:21.980 or Cursor. But for the non -technical founder, 00:04:22.180 --> 00:04:24.879 they just need a custom chat app, something where 00:04:24.879 --> 00:04:27.360 they can ask plain language questions without 00:04:27.360 --> 00:04:29.980 learning complex developer tools. Make it as 00:04:29.980 --> 00:04:32.529 simple as texting a colleague. Precisely. But 00:04:32.529 --> 00:04:34.889 let's really dig into that base layer, the context 00:04:34.889 --> 00:04:37.850 OS. I think a lot of people imagine this as some 00:04:37.850 --> 00:04:41.170 highly complex abstract cloud architecture. That's 00:04:41.170 --> 00:04:43.269 totally. But it's actually incredibly literal. 00:04:43.410 --> 00:04:46.170 It is a literal folder on a machine. Just a normal 00:04:46.170 --> 00:04:49.310 folder. Yeah. And it contains exactly six specific 00:04:49.310 --> 00:04:51.550 Markdown files. And just for anyone who isn't 00:04:51.550 --> 00:04:54.410 a developer, Markdown files are simple text files 00:04:54.410 --> 00:04:57.290 with no formatting that AI reads easily. Right. 00:04:57.410 --> 00:05:00.350 There's no weird PDF styling or hidden code. 00:05:00.470 --> 00:05:02.810 It's just. pure text. So what actually goes into 00:05:02.810 --> 00:05:06.490 these six core files? The first is offers .dodge 00:05:06.490 --> 00:05:09.230 .md. That lists your actual products, your pricing 00:05:09.230 --> 00:05:12.250 tiers, and your target audience. The second is 00:05:12.250 --> 00:05:15.069 icp .md. That's your ideal customer profile. 00:05:15.449 --> 00:05:17.509 I love the idea of using a dedicated file just 00:05:17.509 --> 00:05:19.910 for the customer profile, but I read in the sources 00:05:19.910 --> 00:05:22.430 that it also includes anti -profiles. Yeah, it 00:05:22.430 --> 00:05:24.629 does. Why is that specific part so important? 00:05:24.889 --> 00:05:27.930 Oh, it's crucial. An anti -profile tells the 00:05:27.930 --> 00:05:31.790 AI exactly who is not a good fit. Right. It prevents 00:05:31.790 --> 00:05:34.810 the AI from suggesting marketing campaigns or 00:05:34.810 --> 00:05:37.209 sales tactics that attract those nightmare clients 00:05:37.209 --> 00:05:39.810 who will just churn in two months anyway. That 00:05:39.810 --> 00:05:41.889 makes a lot of sense. You're setting hard boundaries. 00:05:42.290 --> 00:05:46.870 What is the third file? Team .md. It lists the 00:05:46.870 --> 00:05:49.269 team members, their roles, and their specific 00:05:49.269 --> 00:05:52.430 approval authorities. Wait, so the AI actually 00:05:52.430 --> 00:05:55.310 knows who has the power to sign off on a discount? 00:05:55.769 --> 00:05:58.430 Exactly. It knows the chain of command. Then 00:05:58.430 --> 00:06:02.430 you have brand .md. That covers preferred vocabulary, 00:06:02.649 --> 00:06:04.730 tone of voice, words the company never uses. 00:06:04.790 --> 00:06:09.589 Got it. The fifth file is history .md. And this 00:06:09.589 --> 00:06:12.759 one is massive. It logs major business decisions 00:06:12.759 --> 00:06:15.860 from the last six to 12 months. Now, why a history 00:06:15.860 --> 00:06:17.740 file? That seems like a bit of overkill for a 00:06:17.740 --> 00:06:19.920 chat AI. Well, think about it. Have you ever 00:06:19.920 --> 00:06:22.500 brainstormed with a brand new employee and they 00:06:22.500 --> 00:06:24.199 pitch an idea you already tried eight months 00:06:24.199 --> 00:06:26.519 ago that failed miserably? Oh, yeah. All the 00:06:26.519 --> 00:06:29.360 time. Right. History .shmd prevents the AI from 00:06:29.360 --> 00:06:31.180 doing exactly that. It knows what you've already 00:06:31.180 --> 00:06:33.860 tried. That's smart. And finally, you have goals 00:06:33.860 --> 00:06:37.620 .md, which outlines the exact revenue targets 00:06:37.620 --> 00:06:40.730 for the current quarter. OK, but... writing a 00:06:40.730 --> 00:06:43.620 cohesive hyper -accurate business plan across 00:06:43.620 --> 00:06:47.300 six text files sounds like a monumental undertaking. 00:06:47.819 --> 00:06:49.459 Most founders don't even have this written down 00:06:49.459 --> 00:06:51.660 for human employees. Right, which is exactly 00:06:51.660 --> 00:06:54.259 why you do not ask them to write it. Oh, OK. 00:06:54.399 --> 00:06:57.399 As the agency, you execute this incredibly fast. 00:06:57.720 --> 00:07:00.759 You sit down and run a recorded 90 -minute interview 00:07:00.759 --> 00:07:03.040 with the founder. It's just a conversation. You're 00:07:03.040 --> 00:07:04.560 literally just asking them about their business. 00:07:04.779 --> 00:07:06.920 Exactly. You ask them about their offers, their 00:07:06.920 --> 00:07:09.480 worst clients, their team dynamics. You take 00:07:09.480 --> 00:07:12.399 that raw transcript. plug it into Claude, and 00:07:12.399 --> 00:07:14.660 prompt it to generate those six markdown files 00:07:14.660 --> 00:07:17.500 using the founder's exact spoken words. Wait, 00:07:17.500 --> 00:07:19.139 you don't rewrite it to sound more professional 00:07:19.139 --> 00:07:22.240 or polished? Absolutely not. The AI needs the 00:07:22.240 --> 00:07:25.579 founder's real, unfiltered language to accurately 00:07:25.579 --> 00:07:28.160 match their voice later on. If they say a certain 00:07:28.160 --> 00:07:30.480 marketing strategy sucked, you want the words 00:07:30.480 --> 00:07:33.779 sucked in that file. That is a brilliantly efficient 00:07:33.779 --> 00:07:36.660 use of time. It is, but there is a very strict 00:07:36.660 --> 00:07:39.040 rule of thumb you absolutely must follow here. 00:07:39.399 --> 00:07:42.220 What's that? keep each of these six files under 00:07:42.220 --> 00:07:46.000 500 to 800 words. I mean, I still wrestle with 00:07:46.000 --> 00:07:48.819 prompt drift myself. So keeping those files strictly 00:07:48.819 --> 00:07:51.160 under 800 words makes total sense. You really 00:07:51.160 --> 00:07:53.360 have to keep the AI highly focused or it starts 00:07:53.360 --> 00:07:55.699 hallucinating. Exactly. It all comes down to 00:07:55.699 --> 00:07:57.819 token limits. Right. And tokens are basically 00:07:57.819 --> 00:08:00.519 the AI's short -term memory limit for reading 00:08:00.519 --> 00:08:04.199 text. Exactly. Now, Claude code has a massive 00:08:04.199 --> 00:08:08.519 context window of 200 ,000 tokens. If your six 00:08:08.519 --> 00:08:11.439 files are concise, your whole Context OS only 00:08:11.439 --> 00:08:14.500 uses about 5 ,000 tokens. Leaving the vast majority 00:08:14.500 --> 00:08:17.579 of the AI's brain power wide open for the actual 00:08:17.579 --> 00:08:20.540 daily work. You've got it. So if a founder just 00:08:20.540 --> 00:08:23.100 dumps a massive 50 -page business plan into that 00:08:23.100 --> 00:08:26.540 folder, what breaks? The AI gets overwhelmed, 00:08:27.000 --> 00:08:29.339 ignores details, and reverts back to generic 00:08:29.339 --> 00:08:32.360 answers. Okay, so a static folder of text files 00:08:32.360 --> 00:08:35.059 is great for background. It knows who the company 00:08:35.059 --> 00:08:37.320 is and what it sounds like. But an operating 00:08:37.320 --> 00:08:40.259 system needs live breathing data to be truly 00:08:40.259 --> 00:08:42.929 useful. If it doesn't know how much money came 00:08:42.929 --> 00:08:45.269 in yesterday, it can't really help you today. 00:08:45.590 --> 00:08:48.230 Right. And that is exactly where layer two, the 00:08:48.230 --> 00:08:50.450 integrations, becomes absolutely critical. This 00:08:50.450 --> 00:08:52.730 is the plumbing. OK. You need to connect five 00:08:52.730 --> 00:08:55.669 core tools, and you have to do it in a very specific 00:08:55.669 --> 00:08:57.549 sequence. Walk us through that sequence. Why 00:08:57.549 --> 00:08:59.990 does the order actually matter? Well, step one 00:08:59.990 --> 00:09:03.350 is always your revenue. Always. You connect Stripe 00:09:03.350 --> 00:09:05.470 or whatever payment processor they happen to 00:09:05.470 --> 00:09:08.590 use. You use a restricted API key with strict 00:09:08.590 --> 00:09:10.919 read -only access. So the AI can see them. money, 00:09:10.980 --> 00:09:13.580 but it absolutely cannot move it. Safety first. 00:09:13.759 --> 00:09:16.820 Always. You can't compromise on that. Step two 00:09:16.820 --> 00:09:20.179 connects the CRM, HubSpot, PipeDrive, whatever 00:09:20.179 --> 00:09:21.940 they already use. That's where the leads live. 00:09:22.480 --> 00:09:25.679 Okay. Step three is marketing data. So meta ads, 00:09:25.899 --> 00:09:28.039 Google Analytics. Hold on. I want to pause right 00:09:28.039 --> 00:09:30.179 there. You mentioned connecting whatever CRM 00:09:30.179 --> 00:09:32.100 they already use. So you aren't forcing them 00:09:32.100 --> 00:09:35.019 to migrate to some entirely new system. Never. 00:09:35.299 --> 00:09:37.500 The whole point of the AIOS is that it layers 00:09:37.500 --> 00:09:40.039 perfectly over their existing chaotic stack. 00:09:40.279 --> 00:09:42.519 you aren't causing more disruption. Got it. That 00:09:42.519 --> 00:09:45.240 lowers the barrier to entry significantly. Okay, 00:09:45.399 --> 00:09:47.980 what are steps four and five? Step four connects 00:09:47.980 --> 00:09:52.179 daily communication, Slack, Gmail, and step five 00:09:52.179 --> 00:09:55.080 connects internal knowledge bases like Notion 00:09:55.080 --> 00:09:58.860 or Google Drive. This all sounds incredibly powerful. 00:09:59.519 --> 00:10:01.940 But if you have ever run a business, you know 00:10:01.940 --> 00:10:05.120 that live dashboards constantly clash with each 00:10:05.120 --> 00:10:07.700 other. Oh, it's a nightmare. Stripe might accurately 00:10:07.700 --> 00:10:11.299 show 28 new paying customers for the week. Meanwhile, 00:10:11.620 --> 00:10:14.340 the MetaAds dashboard proudly claims it got you 00:10:14.340 --> 00:10:17.700 40 conversions. Yes, the classic dashboard lie. 00:10:17.820 --> 00:10:19.679 Everyone knows it. So if the AI is looking at 00:10:19.679 --> 00:10:22.860 both of those sources. Does it just average them 00:10:22.860 --> 00:10:25.379 out? Does it take a guess? No, that would be 00:10:25.379 --> 00:10:27.919 disastrous. This is exactly why you have to create 00:10:27.919 --> 00:10:30.779 one additional absolutely crucial file in your 00:10:30.779 --> 00:10:32.840 context OS. OK, what is it called? It's called 00:10:32.840 --> 00:10:36.080 datatrust .md. OK, and what exactly does that 00:10:36.080 --> 00:10:39.179 do? It acts as a rigid hierarchy. It establishes 00:10:39.179 --> 00:10:42.220 absolute ground truth for the AI. You explicitly 00:10:42.220 --> 00:10:45.240 write rules telling the AI, for example, Stripe 00:10:45.240 --> 00:10:47.899 is the ultimate undeniable truth for revenue. 00:10:48.720 --> 00:10:51.440 MetaAds is a secondary source that frequently 00:10:51.440 --> 00:10:54.340 overreports, and its numbers always require a 00:10:54.340 --> 00:10:57.720 strict cross -check against Stripe. Whoa! I mean, 00:10:58.080 --> 00:11:01.200 imagine an AI instantly cross -referencing chaotic 00:11:01.200 --> 00:11:04.320 ad dashboards with actual Stripe revenue. and 00:11:04.320 --> 00:11:07.259 doing it without opening a single tab. It's beautiful. 00:11:07.559 --> 00:11:10.379 So when the AI evaluates a campaign, it sees 00:11:10.379 --> 00:11:13.139 Meta claiming 40, but it's hard -coded to verify 00:11:13.139 --> 00:11:15.980 with Stripe. Right. It sees 28 and calculates 00:11:15.980 --> 00:11:18.179 the real return on investment based solely on 00:11:18.179 --> 00:11:21.299 the 28. That completely eliminates hours of manual 00:11:21.299 --> 00:11:23.320 spreadsheet reconciliation for the founder. It 00:11:23.320 --> 00:11:26.039 saves them so much pain. Why is Stripe specifically 00:11:26.039 --> 00:11:28.620 step one in the integration order before leads 00:11:28.620 --> 00:11:30.860 or marketing? Because the AI must see real money 00:11:30.860 --> 00:11:33.799 first to make accurate business decisions. sponsor. 00:11:34.820 --> 00:11:36.820 Welcome back. So we have this foundation successfully 00:11:36.820 --> 00:11:39.539 built. The Context OS knows the business deeply. 00:11:40.080 --> 00:11:42.600 Live data is actively flowing in through the 00:11:42.600 --> 00:11:44.879 integrations. Yep, the pipes are working. But 00:11:44.879 --> 00:11:47.639 we don't just hand this massive system over to 00:11:47.639 --> 00:11:51.059 the client and say, hey, have fun. No, that is 00:11:51.059 --> 00:11:53.980 exactly how you get a confused client who churns 00:11:53.980 --> 00:11:56.779 in a month. Right. We have to find the one specific 00:11:56.779 --> 00:12:00.259 problem that pays for the entire system immediately. 00:12:00.519 --> 00:12:02.820 How do you uncover that specific problem? You 00:12:02.820 --> 00:12:05.200 run a clear discovery session with the founder. 00:12:05.860 --> 00:12:08.639 You ask four beautifully simple questions to 00:12:08.639 --> 00:12:10.659 surface the real pain they're feeling. Give me 00:12:10.659 --> 00:12:13.500 an example. First one. What task did you do this 00:12:13.500 --> 00:12:16.139 week that you also did last week? Right. Looking 00:12:16.139 --> 00:12:19.039 for repetitive manual labor. What else? Which 00:12:19.039 --> 00:12:21.580 report do you desperately wish you had every 00:12:21.580 --> 00:12:23.179 Monday morning, but you never have the time to 00:12:23.179 --> 00:12:25.659 actually build it? That's a good one. Or where 00:12:25.659 --> 00:12:28.539 do valuable leads or customers constantly slip 00:12:28.539 --> 00:12:31.600 through the cracks? And finally, which administrative 00:12:31.600 --> 00:12:34.080 task takes two full hours every single week? 00:12:34.460 --> 00:12:36.519 So they give you a list of their daily headaches. 00:12:37.179 --> 00:12:38.620 Then what do you do with that? You take their 00:12:38.620 --> 00:12:40.960 answers and you score them. You score based on 00:12:40.960 --> 00:12:44.899 two metrics, total hour saved and direct revenue 00:12:44.899 --> 00:12:47.399 impact. The problem with the highest combined 00:12:47.399 --> 00:12:50.559 score becomes your very first build. It's the 00:12:50.559 --> 00:12:53.620 quick win. So instead of building random automations, 00:12:53.740 --> 00:12:56.980 you build highly targeted solutions. And to do 00:12:56.980 --> 00:12:59.419 that, you write very specific prompts into the 00:12:59.419 --> 00:13:02.139 system. But these aren't just simple one -liners, 00:13:02.299 --> 00:13:05.379 right? Far from it. AIS prompts are essentially 00:13:05.379 --> 00:13:08.860 mini briefs. They detail the exact goal, the 00:13:08.860 --> 00:13:11.100 strict limitations, the specific data inputs 00:13:11.100 --> 00:13:13.700 required, and the exact format the output needs 00:13:13.700 --> 00:13:15.850 to take. Let's make this concrete for the listeners. 00:13:16.049 --> 00:13:18.250 Give me a real -world example of a prompt that 00:13:18.250 --> 00:13:20.610 instantly pays for the system. Okay, let's look 00:13:20.610 --> 00:13:23.429 at a weekly ad review prompt. This is incredibly 00:13:23.429 --> 00:13:25.870 powerful for e -commerce brands. Okay, walk me 00:13:25.870 --> 00:13:28.110 through it. You write a prompt where the AI reads 00:13:28.110 --> 00:13:32.110 the offers .dash .md file and the icp .md file. 00:13:32.190 --> 00:13:34.669 It pulls the last seven days of meta ads data. 00:13:35.210 --> 00:13:37.570 But crucially, it cross -checks that against 00:13:37.570 --> 00:13:39.870 the Stripe CSV, the ground truth we established 00:13:39.870 --> 00:13:43.350 earlier. Following the exact datatrust .md rules. 00:13:43.470 --> 00:13:46.529 Exactly. The prompt asks the AI to find the three 00:13:46.529 --> 00:13:49.289 specific ad creatives with the lowest real cost 00:13:49.289 --> 00:13:52.049 per lead based only on Stripe revenue. So it 00:13:52.049 --> 00:13:54.149 completely ignores the inflated meta metrics. 00:13:54.429 --> 00:13:57.889 Right. This specific prompt can expose a hidden 00:13:57.889 --> 00:14:02.029 28 % tracking gap between what meta claims and 00:14:02.029 --> 00:14:05.649 what actually happened. Wow. It identifies exactly 00:14:05.649 --> 00:14:08.269 where precious ad budget is being completely 00:14:08.269 --> 00:14:10.870 wasted. I mean, if you're spending 10 grand a 00:14:10.870 --> 00:14:15.190 month on ads, exposing a 28 % gap pays for the 00:14:15.190 --> 00:14:18.200 agency fee instantly. In week one, it's undeniable. 00:14:18.299 --> 00:14:21.019 Let's look at a second prime example, the sales 00:14:21.019 --> 00:14:24.059 call prep prompt. Oh, I love this one. It's brilliant. 00:14:24.240 --> 00:14:26.860 The prompt tells the AI to prepare the founder 00:14:26.860 --> 00:14:29.960 for an upcoming prospect call. It dives into 00:14:29.960 --> 00:14:32.639 the CRM, pulls the specific prospect record, 00:14:32.720 --> 00:14:35.620 and then it actually searches the connected Gmail 00:14:35.620 --> 00:14:38.659 for the last three email threads with that exact 00:14:38.659 --> 00:14:40.539 person. So instead of the founder frantically 00:14:40.539 --> 00:14:42.559 searching their inbox five minutes before the 00:14:42.559 --> 00:14:45.740 Zoom call begins. The AI does it, and it seamlessly 00:14:45.870 --> 00:14:48.629 outputs a tight two -line summary of the relationship. 00:14:48.970 --> 00:14:51.389 It predicts the prospect's three most likely 00:14:51.389 --> 00:14:53.909 objections based on past data. That's incredible. 00:14:54.029 --> 00:14:56.029 And it provides the perfect opening question 00:14:56.029 --> 00:14:58.509 to start the call. And this just lands in the 00:14:58.509 --> 00:15:00.870 founder's inbox exactly one hour before the meeting. 00:15:01.070 --> 00:15:02.809 Automatically. They don't even have to ask for 00:15:02.809 --> 00:15:06.490 it. And over time, you build five to 10 of these 00:15:06.490 --> 00:15:09.990 reusable prompts. OK. They become the actual 00:15:09.990 --> 00:15:12.250 product. They become the very thing the founder 00:15:12.250 --> 00:15:14.960 touches and relies on every single week. That 00:15:14.960 --> 00:15:17.600 small library of tailored solutions is really 00:15:17.600 --> 00:15:20.080 what keeps the client happily paying month after 00:15:20.080 --> 00:15:23.659 month. Why is exposing that 28 % tracking gap 00:15:23.659 --> 00:15:26.700 the ultimate first move for an agency? It delivers 00:15:26.700 --> 00:15:29.419 instant financial ROI, making the initial setup 00:15:29.419 --> 00:15:32.340 fee an easy yes. We've successfully delivered 00:15:32.340 --> 00:15:35.639 that undeniable quick win. Now, how does the 00:15:35.639 --> 00:15:38.039 agency actually structure the underlying business 00:15:38.039 --> 00:15:40.620 model? Right. How do you totally avoid that $1 00:15:40.620 --> 00:15:42.840 ,500 burnout cycle we discuss at the very top 00:15:42.840 --> 00:15:45.149 of the show? When you need a clear position on 00:15:45.149 --> 00:15:48.210 the map, the sources lay out three specific offer 00:15:48.210 --> 00:15:49.970 shapes you can choose from. Let's break them 00:15:49.970 --> 00:15:52.309 down. The first shape is the training and setup 00:15:52.309 --> 00:15:55.990 package. You charge a $5 ,000 initial setup fee 00:15:55.990 --> 00:16:00.129 to build the AIOS. Then you offer a short support 00:16:00.129 --> 00:16:02.750 window for maybe $1 ,000 to $2 ,000 a month. 00:16:02.970 --> 00:16:04.990 But there is a massive risk here, right? Yeah, 00:16:05.210 --> 00:16:07.710 a very high risk of the founder just leaving. 00:16:08.289 --> 00:16:10.250 After three months, they learn how the system 00:16:10.250 --> 00:16:12.730 works, they realize they don't need you to babysit 00:16:12.730 --> 00:16:15.350 it, and they just run it themselves. You are 00:16:15.350 --> 00:16:17.649 right back to hunting for new clients to survive. 00:16:17.870 --> 00:16:20.429 So what is the second shape? It's a highly productized 00:16:20.429 --> 00:16:24.370 AIS. You pick one single hyper -specific niche, 00:16:24.669 --> 00:16:28.009 say... roofing companies in Texas. You build 00:16:28.009 --> 00:16:31.309 a highly targeted version of the context OS just 00:16:31.309 --> 00:16:34.250 for them. You charge roughly $10 ,000 for the 00:16:34.250 --> 00:16:36.629 setup. That model seems like it works wonderfully 00:16:36.629 --> 00:16:39.330 for agencies that want massive, rapid scale. 00:16:39.590 --> 00:16:41.649 Absolutely, because your setup time drops to 00:16:41.649 --> 00:16:43.889 almost zero after your first five builds. You're 00:16:43.889 --> 00:16:46.649 basically just duplicating the exact same architecture. 00:16:46.809 --> 00:16:49.250 But there is a third shape, the retainer with 00:16:49.250 --> 00:16:52.980 install. This is the absolute most useful option 00:16:52.980 --> 00:16:54.200 right now, especially if you're dealing with 00:16:54.200 --> 00:16:56.100 non -technical SMB owners. How does it work? 00:16:56.340 --> 00:16:59.480 You charge a $5 ,000 setup fee upfront to build 00:16:59.480 --> 00:17:02.720 the base layer. Then you charge a monthly retainer 00:17:02.720 --> 00:17:07.680 of $2 ,500 to $6 ,000. Now wait, why does this 00:17:07.680 --> 00:17:10.319 specific retainer model actually work so perfectly 00:17:10.319 --> 00:17:14.079 now in 2026 when it didn't a few years ago? Because 00:17:14.079 --> 00:17:16.859 of massive improvements in sheer build speed. 00:17:17.579 --> 00:17:20.619 The tooling has evolved so much that one solo 00:17:20.619 --> 00:17:22.799 operator can now do the technical work of three 00:17:22.799 --> 00:17:25.779 people. You're relying heavily on tools like 00:17:25.779 --> 00:17:30.220 Claude Code, Cursor, and ICNOT. Quick pause for 00:17:30.220 --> 00:17:32.059 those listening who aren't deeply technical. 00:17:32.579 --> 00:17:34.980 What is NN? Is that essentially just a more advanced 00:17:34.980 --> 00:17:37.960 version of Zapier? Basically, yes. It's an incredibly 00:17:37.960 --> 00:17:40.579 powerful workflow automation tool that lets all 00:17:40.579 --> 00:17:42.359 these different apps talk to each other under 00:17:42.359 --> 00:17:44.720 the hood. Got it. So the agency handles all that 00:17:44.720 --> 00:17:47.440 complex back end wiring. Yeah. But the founder 00:17:47.440 --> 00:17:49.799 still needs extreme front end simplicity, right? 00:17:49.920 --> 00:17:52.980 Yes. Most business owners have zero desire. to 00:17:52.980 --> 00:17:55.819 look at a complex workflow canvas or learn Claude 00:17:55.819 --> 00:17:58.400 code. Right. So you use tools like Lovable or 00:17:58.400 --> 00:18:01.259 V0 by Vercell. They are platforms that let you 00:18:01.259 --> 00:18:03.759 use plain English to rapidly build a custom, 00:18:03.880 --> 00:18:06.799 really clean chat interface. Ah, so you essentially 00:18:06.799 --> 00:18:09.440 build a custom app for the founder. Right. It 00:18:09.440 --> 00:18:11.680 gives them a simple, beautiful interface on their 00:18:11.680 --> 00:18:14.420 phone or laptop that is tied directly into their 00:18:14.420 --> 00:18:17.180 specific context OS. So they just type a normal 00:18:17.180 --> 00:18:19.359 question and your complex backend does all the 00:18:19.359 --> 00:18:22.279 heavy lifting. Exactly. But if you're on a monthly 00:18:22.279 --> 00:18:24.700 retainer, you have to fiercely protect your profit 00:18:24.700 --> 00:18:26.900 margin over time. Otherwise, the client will 00:18:26.900 --> 00:18:29.480 just ask for endless tweaks. Scope creep will 00:18:29.480 --> 00:18:32.259 utterly destroy your agency if you let it. You 00:18:32.259 --> 00:18:35.200 have to use a strictly defined four -line contract. 00:18:35.339 --> 00:18:38.400 Only four lines. Keep it simple. Line one, you 00:18:38.400 --> 00:18:41.480 guarantee exactly one major automation per month, 00:18:41.700 --> 00:18:44.259 defined and signed off in advance. Okay. Line 00:18:44.259 --> 00:18:48.039 two, you allow up to two small edits or bug fixes 00:18:48.039 --> 00:18:51.890 to existing automations. Line three, you include 00:18:51.890 --> 00:18:55.049 one 30 -minute strategy call weekly. And line 00:18:55.049 --> 00:18:58.869 four, you provide one monthly ROI report explicitly 00:18:58.869 --> 00:19:01.430 proving your financial value. That is beautifully 00:19:01.430 --> 00:19:04.190 contained. So anything outside that specific 00:19:04.190 --> 00:19:06.250 scope, say they want a massive new marketing 00:19:06.250 --> 00:19:09.089 workflow mid -month, that becomes a completely 00:19:09.089 --> 00:19:11.930 separate paid project. Exactly. You set that 00:19:11.930 --> 00:19:14.450 hard rule perfectly clearly on day one. Won't 00:19:14.450 --> 00:19:18.279 clients push back? if you rigidly limit them 00:19:18.279 --> 00:19:20.779 to just one major automation a month. No. Writing 00:19:20.779 --> 00:19:23.240 the rules clearly on day one actually earns the 00:19:23.240 --> 00:19:25.279 founders respect. We have covered a tremendous 00:19:25.279 --> 00:19:28.000 amount of ground today. Yeah. Let's slowly zoom 00:19:28.000 --> 00:19:30.539 out for the big idea here. The era of selling 00:19:30.539 --> 00:19:34.019 disposable $1 ,500 chatbots is officially over. 00:19:34.240 --> 00:19:37.779 Business owners do not need more random disconnected 00:19:37.779 --> 00:19:39.980 tools. They don't need another tab open. No, 00:19:39.980 --> 00:19:42.319 they don't. They desperately need an intelligence 00:19:42.319 --> 00:19:45.500 foundation that deeply, permanently understands 00:19:45.500 --> 00:19:47.599 their business. You are no longer selling tiny, 00:19:47.599 --> 00:19:49.640 isolated automations. You are effectively selling 00:19:49.640 --> 00:19:52.160 an entire intelligence foundation. Right. And 00:19:52.160 --> 00:19:55.279 here's the secret. When you fully own that context 00:19:55.279 --> 00:19:58.660 OS, when you manage the digital memory, you own 00:19:58.660 --> 00:20:00.900 the long -term relationship. You successfully 00:20:00.900 --> 00:20:03.599 shift from being a completely replaceable freelancer 00:20:03.829 --> 00:20:07.210 to an indispensable partner. And as the underlying 00:20:07.210 --> 00:20:09.869 AI technology continues to get faster and cheaper, 00:20:10.650 --> 00:20:12.789 your profit margins on that retainer will actually 00:20:12.789 --> 00:20:15.789 increase. Your manual effort steadily decreases 00:20:15.789 --> 00:20:18.609 over time. Stop selling temporary fixes. Start 00:20:18.609 --> 00:20:20.890 building their future operating system. I want 00:20:20.890 --> 00:20:23.369 to leave you with one final pretty provocative 00:20:23.369 --> 00:20:26.269 thought today. Think back to those Lego blocks 00:20:26.269 --> 00:20:29.339 of data we mentioned earlier. If you carefully 00:20:29.339 --> 00:20:32.140 stack those blocks and you execute this retainer 00:20:32.140 --> 00:20:35.119 model perfectly over a year or two, things change. 00:20:35.240 --> 00:20:38.640 It naturally leads to the very top of the agency 00:20:38.640 --> 00:20:42.400 pyramid. Oh, the equity play. Exactly. Imagine 00:20:42.400 --> 00:20:45.220 not just passively collecting a flat monthly 00:20:45.220 --> 00:20:48.000 fee from your clients. Imagine actively helping 00:20:48.000 --> 00:20:51.359 launch brand new AI first businesses from scratch. 00:20:51.539 --> 00:20:53.799 Right. You could strategically trade your deep 00:20:53.799 --> 00:20:56.740 context OS expertise for actual company equity. 00:20:57.000 --> 00:20:59.440 You could negotiate long term profit share agreements 00:20:59.440 --> 00:21:01.759 instead of just retainer fees. That is truly 00:21:01.759 --> 00:21:04.299 the ultimate end game for a smart agency owner. 00:21:04.420 --> 00:21:06.779 You become a true partner in their growth. So 00:21:06.779 --> 00:21:08.950 here is is your clear challenge for this upcoming 00:21:08.950 --> 00:21:12.109 week. Don't just passively absorb this deep dive 00:21:12.109 --> 00:21:15.009 and move on. Go find one business, and maybe 00:21:15.009 --> 00:21:18.069 even your own, struggling heavily with chaotic 00:21:18.069 --> 00:21:20.809 tool sprawl. Yeah, find that pain point. Run 00:21:20.809 --> 00:21:23.470 a targeted 90 -minute session with that specific 00:21:23.470 --> 00:21:26.650 founder, build those six essential markdown files, 00:21:27.069 --> 00:21:29.410 and establish their base layer. Take decisive 00:21:29.410 --> 00:21:32.130 action. Prove the concept and secure that immediate 00:21:32.130 --> 00:21:34.430 financial return. Thank you so much for joining 00:21:34.430 --> 00:21:36.650 our deep dive today. Take good care of yourselves 00:21:36.650 --> 00:21:38.750 out there. We will catch you on the next one. 00:21:38.970 --> 00:21:39.789 Out to your own music.