WEBVTT

00:00:00.000 --> 00:00:03.680
A few weeks ago, I caught myself doing something weird.

00:00:03.680 --> 00:00:10.880
It was Sunday night, my laptop was open, and Claude was researching the latest AI market updates. And then I watched it open NotebookLM,

00:00:10.880 --> 00:00:37.940
drop in everything it found, and click audio overview. Five minutes later, I had a podcast about my own market sitting on my computer. I did not write it, I did not record it, and I didn't even stay awake for it. The work had happened, and I just wasn't doing it. That is the line that most people never cross with AI. They use it like a consumer. They open the app, they ask a question, they close the tab. Day after day, year after year, they got faster, but they didn't get free. Every founder that I know uses Cloud. Every founder also uses NotebookLM,

00:00:37.940 --> 00:01:10.540
but almost nobody connects them. And the ones who do are running an entire engine that most people are still doing by hand. Thank you to HubSpot for partnering with us on today's video. I shared this with one of my communities and they went totally crazy. So I decided to share it with you here as well. I've built three of these chains and I will show you all three. We're gonna build the last one together, twenty minutes to set up. It runs every week forever and I think that's the trait. Twenty minutes once versus doing this by hand for the rest of your career. So let's get into it. There's a McKinsey study that found that knowledge workers lose nearly 20%

00:01:10.540 --> 00:01:17.725
of their week just searching for and gathering information. To To me, that is totally mind blowing, and I need you to read it again. 20%,

00:01:17.725 --> 00:02:06.965
one full day every week is gone digging through files and tabs and tools hunting for something that you already know exists. That is the tax that you pay when your tools live in silos, and almost everyone is paying it without noticing. Here's why. You were trained to think of AI as a stack of better apps. ChatGPT for this and Claude for that and NotebookLM for this other thing. You open, you use, you close, you repeat. I think that is the consumer pattern. The consumer pattern has a CV. Think about it this way. Using notebook l m without clot is like hiring a brilliant analyst and making them sit in a room with no Internet, and they can only do incredible work when you remember to bring them something or you open the door. They never go out and find anything on their own. They never follow-up. They just sit there waiting for you. That's Notepo KLM alone. It's brilliant. It's grounded. It's source cited, but it's also stuck. Claude

00:02:06.965 --> 00:02:13.845
has the opposite problem. It can go anywhere. It can research. It can automate. It can browse. It can deliver, but it can't make audio overviews.

00:02:13.845 --> 00:02:22.800
It cannot build source grounded mind maps. It cannot generate video overviews, and working from general knowledge instead of your specific documents, it hallucinates

00:02:22.800 --> 00:02:54.240
many times. One can't move, the other can't think. Apart their apps, but together, they can become an engine. So I wanna show you this blueprint. It has three chains. Each one kills a different problem that you have been hand grinding for years, I'm pretty sure, because a lot of people have. Chain number one is the autopilot brief. Cloud researches your prospect. It feeds it into NotebookLM. You walk into the call looking like you've been studying them for weeks. The move inside is the auto source feed. Chain number two is the auto refresh loop. So Cloud watches your support channels for edge cases.

00:02:54.400 --> 00:02:55.200
NotebookLM

00:02:55.200 --> 00:03:08.645
regrounds your agent so it never goes stale. The move inside is the edge case sweep. I'm gonna come back and explain everything. Chain number three is the competitive radar. So, basically, this one is where Cloud researches your competitors every week. NotebookLM

00:03:08.645 --> 00:03:38.095
turns it into a source grounded podcast waiting on your computer or your phone Monday morning. The output is the morning briefing. There's three chains, one engine, each one runs without you in the room. So let me explain how to build everything. So first chain. This one is about closing clients. Who doesn't want that? Right? Most founders prepare for prospect calls the same way. The call is in thirty minutes. You panic open the website. You skim up their about page. You glance at their LinkedIn, maybe read one post, and then walk into the call and improvise.

00:03:38.460 --> 00:05:00.310
And honestly, it usually works well enough, but well enough leaves money on the table. The founders who win the deal are not the ones who know the most. They're the ones who make the prospect feel understood before a single word is spoken. So here's how the autopilot brief AI engine can do that differently. So let's say a prospect books a call. You can tell Claude one thing. You can say research the prospect name, pull their website, their recent LinkedIn posts, any press from the last ninety days, then open my client Intel notebook in notebook l m and add what you found. This is the auto source feed. Claude goes out into the world, gathers everything, and delivers it directly to Notebook LM. There's no copy pasting, no tab switching from your side. Claude operates the browser and adds the sources on its own. Now your notebook has everything. The prospect's world combined with your case studies that were already loaded, and then you walk over and click two buttons. Mind map to see where your experience overlaps with their problems, and video overview to generate a personalized briefing that you can send before the call. Now the research that used to take fifteen minutes, Claude is able to do in probably two. The synthesis, NotebookLM can do it maybe in one or two as well, and you spent thirty seconds clicking two buttons. Look. I'll tell you when this really clicked for me. Last month, I was working with a consulting client. We had a call recording from our discovery sessions as well as our pricing strategy document already sitting in NotebookLM.

00:05:00.310 --> 00:05:18.625
And Claude went out and pulled their latest quarterly results, added them directly as a new source, and then we did something that I hadn't tried before. We generated an audio overview, the interactive kind actually that you can talk to, and then we brainstormed with it out loud. We walked around the office, challenged the recommendations, pressure tested the pricing.

00:05:18.785 --> 00:05:26.090
It was like having a third person in the room who had actually read everything and never forgot a detail. By the time that we sat down to write proposal,

00:05:26.170 --> 00:06:03.570
the thinking was already done, and we got a lot of help from Claude after that as well. Now look, basically, what had happened there is that Claude did the legwork, and NotebookLM did the thinking. We just showed up and had our personal point of view. So that's the system. But if you're watching this and thinking, I don't wanna build this from scratch or spent hours rewriting prompts or trying to get the research instructions right. Well, HubSpot actually put together something that gets you most of the way there. It is called the AI sales agent kit, and it's three plug and play agents that you can drop straight into clot. The first one defines your ideal customer profile using your actual data, so Cloud knows which prospects are worth researching in the first place.

00:06:03.810 --> 00:06:42.385
The second one qualifies accounts and tells you which leads deserve your time, and the third is basically a prebuilt version of what I just showed you. All you need to do is point it at a prospect and get a full briefing before the call. Now you don't have to use all three, but if you want a starting point instead of writing every prompt from scratch, this is going to save you the trial and error. Hit the link below to grab the AI sales agent kit. It is a 100% free. And by the way, big thanks to HubSpot for partnering with us on today's video and making this resource available. Now here's what's interesting. The prospect doesn't know any of this happened. They just think that you're unusually well prepared, and that's what good systems do. They make the output look effortless.

00:06:42.465 --> 00:06:51.025
Now there's one rule though. Never send raw output. Okay? Always skim the audio or the video overview. Tighten the tone. Cut anything generic.

00:06:51.185 --> 00:06:58.850
The engine is able to manufacture the assets, but you're still the quality control, if you get what I mean. Now the second chain, this one is

00:06:59.010 --> 00:06:59.890
quieter,

00:07:00.130 --> 00:08:38.085
but it's also probably going to matter more in the mid to long term. So if you're building AI automations for clients, I don't know, support agents or onboarding flows or sales bots, you already know what happens at month three. The product changes, new features ship, prices get updated, but the agent doesn't know. It's still confidently answering questions based on documents that you uploaded ninety days ago. And trust me, a stale agent is like a salesperson who stopped reading the product updates six months ago. They're still confident, but they're just wrong. And confident but wrong is worse than not answering at all because the client trusts the answer until they don't, and then you lose the retainer. And the irony is that most founders who build AI agents don't use AI to maintain them. They build the future and then manage it with spreadsheets and calendar reminders. I mean, don't do that. Okay? Promise me. Here's the auto refresh loop, and hopefully, you're gonna implement and use this one. The key or ninja move inside this one, I call the edge case sweep. So step number one, you build your agent's knowledge base inside NotebookLM. Your SOPs, your product documents, your past support tickets, offer pages, whatever you need. NotebookLM is source grounded by design. So every answer cites a document. No hallucinations are going to happen. This is your agent's foundation. Step number two, you set up Claude to run the edge case sweep. Every week, Claude scans your support inbox, your Slack channels, your help desk. It's going to look for three things, questions that the agent couldn't answer, complaints about outdated information, and new product features mentioned by the team that have not made it into the documents yet. And step number three, Claude takes those findings and adds them as new sources to the agent's NotebookLM notebook. And NotebookLM regrounds the entire knowledge base automatically.

00:08:38.245 --> 00:08:48.165
You export the updated briefing and feed it back into the agent. This way, your agent's knowledge base now maintains itself. New edge cases get absorbed, product updates get integrated,

00:08:48.325 --> 00:09:20.750
the agent doesn't go stale because the edge key sweep runs whether you remember it or not. Now there is again one rule, review clause additions before they go live. The loop is automated, but the judgment call has to still be yours. Now I think this is going to be the difference between an AI agency that churns clients at month three and one that keeps them into, I don't know, year one, year two, and beyond. And if you've been in this business long enough, you know that retention is where the margins actually live, not in the first invoice, in the twelfth. Alright. Now let's talk about chain number three. This one is the competitive rater.

00:09:20.910 --> 00:09:23.710
Every week, Claude researches your top competitors,

00:09:23.790 --> 00:09:27.150
product updates, pricing changes, new content, press mentions.

00:09:27.390 --> 00:09:40.555
It feeds those findings into a NotebookLM notebook, and NotebookLM generates a source grounded mind map and an audio overview. The result is what I call your Monday briefing, a competitive intelligence podcast waiting for you when your week starts.

00:09:40.795 --> 00:09:58.020
Now running a business without a competitive intelligence is like driving with your mirrors covered. You can still move forward, but you just can't see what's coming up behind you. The Monday briefing is going to fix the mirrors. So let me explain how to build it. Part number one, you need to build the brain. Okay? So you open NotebookLM.

00:09:58.020 --> 00:10:41.115
You create a new notebook. You can call it, I don't know, competitive radar or whatever you like. And then you add your starting sources, your competitors' websites, their pricing pages, your own positioning document, and this is going to give NotebookLM the context that it needs to spot what changed. It just takes two minutes, and you only have to do it once. Okay? Now part number two, you need to give it legs, and that's where Claude becomes important. So Claude with the Chrome extension is what's going to connect the two. Alright? So that's what gives Claude the ability to open notebook l m and interact with it directly. We don't use APIs. We don't use code. Claude just operates the browser the same way that you would accept. It does it at two in the morning while you sleep if you leave your computer open. Part number three, you

00:10:40.950 --> 00:11:09.405
instructions. So every engine needs a set of instructions. This is the prompt that tells Claude what to do each cycle. Now I'm gonna give you the structure, and the exact templates are going to be inside of our community. But here's the gist. You need to say, research the latest updates from competitor a and competitor b and competitor c. Please look for product launches, pricing changes, new features, and press from the last seven days. Open my competitive radar notebook in NotebookLM and add the most important findings as sources.

00:11:09.725 --> 00:11:20.980
Then generate a fresh mind map, and I want you to look at what happens. Claude is going to do the research, pull all the latest information from across the web, and then it's going to open NotebookLM,

00:11:21.140 --> 00:11:24.500
and you can see it navigating to your notebook.

00:11:24.580 --> 00:11:37.215
And you can see it adding the new sources, And then you have a mind map, and that mind map is going to rebuild itself with the weak's competitive intelligence. It's source grounded. Every claim is cited. There's no hallucinations.

00:11:37.215 --> 00:12:46.760
You didn't touch it. Okay? I think this is what makes this combination an engine instead of a tool because you set it up as a scheduled task. I'm running mine every Sunday at 8PM, but you can choose whenever you wanna run yours. And Claude executes the full cycle automatically. Research, feed, generate every single week. And then Monday morning, you open Notebook LM, and a fresh mind map with this week's intelligence is going to wait for you. And all you need to do is click audio overview and generate your Monday briefing. It's gonna take a few minutes. It's gonna be source cited and ready for your commute. So you are now briefed on everything that your competitors did the past week. You didn't open a tab. You didn't copy a link. You didn't do any research. The engine ran everything while you slept. It took twenty minutes to build, but it'll run forever. That is the competitive radar. Now let me take a step back and talk about what these three chains actually mean for your business. The autopilot brief. You can think of it this way. Prospects who receive a personalized pre call briefing show up prepared and pre sold. Close rates move because trust is built before the first handshake, and the auto source feed eliminated the prep time that used to make you skip this entire step when things got busy. The deals you lose because you showed up underprepared,

00:12:46.920 --> 00:13:28.495
those are the ones that you never even knew you lost. Now the auto refresh loop, I cannot even begin to explain how important this is. Grounded agents that stay current keep clients longer. The Edge k Sweep makes long retention possible without you manually checking docs every week. And if you run an agency, you already know that the real revenue isn't in signing clients, it's in keeping them. Now the repetitive radar. Founders who know what shifted in their market last week, especially in AI, make better decisions about product positioning pricing. The ones who don't are reacting instead of leading, but you can have your Monday briefing to make a difference. Now, of course, results are going to depend on your niche, on your offer, on your consistency in running the engines.

00:13:28.735 --> 00:14:02.420
This is not a shortcut. It's It's an infrastructure, and infrastructure can compound, but only if you build it and let it run. Now let me bring this full circle. Okay? A lot of people, even here on YouTube, a lot of creators are teaching you AI tools at a time. Here's Clone. Here's NotebookLM. Pick your favorite. Watch the tutorial. You now know something that they're not showing you. The leverage is not in the tool. It's in the chain, in the workflow, in the way you connect to these. Claude does what NotebookLM can't, and NotebookLM creates what Claude can't. You can connect them now into an AI engine, and you have a system that researches, that synthesizes,

00:14:02.500 --> 00:14:06.340
and that delivers while you're not even at your desk. The real change

00:14:06.580 --> 00:14:10.980
is not about the tool you use. It's about how you think. Consumers

00:14:11.060 --> 00:14:24.445
use apps. Founders build engines. And once you start seeing AI as the infrastructure that it is, I promise you, you are not going to see any way of going back. Now if you want the exact prompts that power these three chains,

00:14:24.880 --> 00:14:30.880
plus the step by step guide, the setup guide, and everything, those live inside the AI business trailblazers.

00:14:30.880 --> 00:14:35.360
Hive, we have done a notebook l m challenge, actually, as well as a quad challenge.

00:14:35.440 --> 00:15:07.805
Members get everything that I share here, and if you want even more support and a ninety day plan, you're always more than welcome to join our founders community as well. Everything is available and linked here and down below. Now if you build one thing from this video, please make it be the competitive radar. Twenty minutes to set it up. It runs every week forever, and you will never be caught off guard to competitors' moves again. Now these three chains are powerful, but they are the foundation. I'm building many many more right now. One of them is for automated content repurposing,

00:15:07.805 --> 00:15:18.365
and that feeds your entire content calendar, and one for client onboarding that runs itself from the moment that someone signs. I'm going to break them down in the upcoming videos, so make sure that you are subscribed.

00:15:18.445 --> 00:15:31.860
And in the meantime, like this video if you did, be sure to subscribe. Like I said, if you haven't done so, share it with anyone in your circle of friends or family or coworkers who you think needs to learn about this AI engine. And until next time, I suggest you go ahead and check this video out, and I'll see you there.