#117 Why Every Company Needs a Second Brain

Show Notes

About 10 days ago, Malcolm met a business CEO at Zurich Airport who explained how he had built the second brain of his company in just 48 hours. That conversation changed everything.

In this episode, Malcolm breaks down what a real company second brain actually is, why most firms still do not have one, and why that is becoming a serious competitive disadvantage. This is not just a chatbot, not just a better SharePoint search, and not just another enterprise AI wrapper. A real second brain continuously ingests company knowledge — emails, CRM data, SharePoint files, financial data, meeting notes, calendars, and more — and turns that into something the business can query, correct, and eventually act through.

Malcolm explains why the missing ingredient was never just a vector database. The breakthrough came from a smarter architecture: a living company memory with a Wikipedia-like intelligence layer on top, plus bi-directional learning so the system can improve when people correct it. That is what turns a static company GPT into something much closer to an actual organizational brain.

He also walks through concrete use cases already happening right now: preparing for customer meetings with far better context, compressing CEO onboarding from months into days, and giving teams access to a searchable memory layer that actually understands customers, projects, risks, invoices, and past work.

The episode then zooms out to the bigger signal. Malcolm connects this directly to SoftBank’s investment thesis and the rise of second brains for robots. The argument is simple: robots need context to operate intelligently, and so do companies. If physical AI is getting a second brain before your employees do, something is off.

At its core, this episode is about leverage. Most companies are still flying blind because their knowledge is fragmented across inboxes, folders, meetings, and disconnected systems. A second brain changes that. And the companies building one now will have a brutal advantage over the ones that wait.

---

🎙️ ABOUT THE HOST

Malcolm Werchota leads AI adoption programs for companies across Europe. After more than 15 years in international corporates and leadership roles, his focus today is practical AI implementation without the usual nonsense.

He works with companies from manufacturing to pharma, from family-owned businesses to large global enterprises — always with a strong bias toward real-world adoption and business value.

---

🚀 RESOURCES FOR LEADERS

📚 Chief AI Academy — AI for Decision-Makers

https://www.werchota.ai/chief-ai-academy

👥 AI Leadership Community

https://chief.werchota.ai/getting-started

---

📬 CONTACT

LinkedIn: https://linkedin.com/in/malcolmwerchota

E-Mail: [e-mail]

---

🔎 TAGS

#AI #AICookbook #SecondBrain #EnterpriseAI #AIAdoption #KnowledgeManagement #MCP #VectorDatabase #CEO #Leadership #Robotics #PhysicalAI #Azure #Supabase #ClaudeCode

What is The AI Cookbook Show by Malcolm Werchota?

Malcolm Werchota's AI Cookbook Show is where artificial intelligence meets authentic business transformation. Known for his direct style and willingness to show AI in action—even during live presentations—Malcolm helps organizations understand that AI isn't about replacing humans but amplifying their capabilities. From voice-note productivity hacks to real-time meeting intelligence, this podcast delivers actionable insights for immediate implementation.

2026_04_19_E117 2nd Brain EN: Okay. So about 10 days ago, Damian and I, walked into Zurich airport. Actually we went to the Starbucks. I don't know if you've been to the Zurich airport, there are two or three Starbucks. And we were at the one that, was the bottom of the train station. And it was a bit like a spy movie. Like it was kind of like, can we meet you? Yes. Let's meet at Zurich airport. Funnily enough, don't know lately, really within the last, I don't know, four, five, six weeks, once a week. I have a meeting at Zurich airport. Switzerland is already pretty neutral, but it seems Zurich airport is even more neutral. I come to Zurich airport and meet me there in a hotel lobby or in the Starbucks. Anyway, so we went there, we met the guy. He's not a developer. Actually, he's the business CEO of a division of a larger company. Guys, he's not a developer. And his division, I think they have like 110 employees, they make 50 to 60 million dollars a year. And he talks to us and he explains to us how he built the second brain of his company. Now, we're trying to build the second brain. I mean, I did a very nice prototype around Christmas and New Year, and then I worked for it for another month. And then I had Thomas, who's a buddy of mine, and he used to be the head of AI of a company. and he kind of looked at this and then he worked on this for another two months and it didn't work. It really didn't really work. And that guy says to me, Malcolm, I built the entire brain of my company in 48 hours. And that really knocked me off my chair because not only have we been spending so much time in trying to do it, but it took a compelling event which is that Andre Capati about two weeks ago explained how to build the second brain. So it took two people three months of failure in our company and this guy over a coffee explains to us how he built it in 48 hours. So obviously because we're know we're builders because we like challenges I went home And four days later, I mean, not 48 hours, but four days later, where Chota AI had its own second brain. And now let's explain what is a second brain. In the second brain is everything, is all our emails from every single employee, is everything that's on SharePoint, is our financial data, is... everything that's in our CRM. This is our second brain. So I'm Malcolm Verrota. This is the bilingual podcast in English. It's called the AI cookbook and in German it's called Das KI Kochbuch. And today we're going to talk about what the second brain exactly is. Three ways that we're already using it and why actually companies are asking us to explain to them how to do it. And not only why we're building it, but why SoftBank, one of the biggest investment fund in the world is pouring billions into a company that is also building second brains but not for you guys, for robots. And I'm thinking to myself, hey, Skilled is building second brains for robots while 99 % of all the companies in the world don't even have a second brain yet. So yeah. And if you listen to the end, will even understand why statistically you who are listening to this podcast is now one out of 70,000 people. You know, when you go into, I don't know, stadium, like for example, the arena, the arena in Munich. why you are one out of all the people in that stadium who actually understands what's going on. So yeah, people, action, let's go. flip the switch the circuits live digital dreams they thrive and strive recipes coded to feed your mind serving knowledge of the AI kind welcome to the AI cookbook show where the data streams and the insights flow ⁓ ⁓ Okay, second brain. Now, if you have been looking and searching the internet a bit, you will see that there are companies that are offering this or offering a part of it. One of them is called Glean. So this is not Glean, okay? Because Glean is nothing else but just a bit of a SharePoint search, right? With a chatbot on top. So it's not a SharePoint search. It's not a chatbot. It's not Glean. The second brain, what we're talking about this concept is three things. Remember consultants speak in three because our brain is very small. So it's every piece of a company data ingested. Now you don't need to go and do it so radically like we did and how many companies are doing it. The way that you can do it is you can just do it already for a project. You can take a project which is running since 10 years. And then you say, okay, I mean, let's get a bit of a better overview of this project. Or you know, some companies have these shared email inboxes where everybody has access to it because it's called, I don't know, invoice at company.com or it's called whatever, know, events at company.com or I don't know what, or for a certain project, this can be ingested. But the beauty is it's not static. So for us, Every night at 2 a.m. our brain sucks in everything new. Everything that happened within the last 24 hours goes into it. Now, where does it go into? Primarily, it goes into a vector database. And anybody who's a bit clever will say, yeah, but we've tried that before, it doesn't work. Yeah, okay. Well, hold your horses. Secondly, what is a second brain? A second brain is querible by everybody. Okay. Who can ask questions to a second brain? Well, everybody who knows the phone number of that second brain. And the second brain, the phone number is called an MCP server. So it's kind of like a phone number and then you can call that company memory. And you can call it from anywhere. If you are using code code, you can call it through code code, using codex from OpenAI, you can call it through codex. You can run a local model like a Gemma 4. This is actually we've updated Now we have three second brains, but that's a separate story. So we've updated all our last second brains with Gemma 4. Gemma 4 is a model from Google. And it's a local model from Google, which means you can use a second brain and you can switch off the internet and it's disconnected from the internet. Like you don't need to be worried that... people are gonna be able to hack your second brain. Or you can host it on Azure because your emails are already on Azure, all your data is on OneDrive on SharePoint. So anyway, it seems you trust Microsoft a lot with all your data. You can build and one of the second brains that we've built is on Azure and it's hosted on Azure. And every employee can ask, hey, I'm about to meet this customer. What do I need to know? And the answer doesn't come like in chat GPT or in cloud in like three minutes, it comes in like 20 minutes. Why? Because it'll get you the timeline. It'll find out all the things that you did for it. It'll find out the project. It'll give you like, ⁓ with this customer, you need to watch out for A and you need to watch out for B and opportunities for this customer as C. And the last thing that makes this brain very cool, it's bi-directional and learning. And that is essentially what Andre Capati taught us. Okay, now what did Andre Capati taught us? Andre Capati said, that was about two weeks ago, he said, look, when you build a second brain for this thing to work very well, RAG we know doesn't really work because RAG is like a huge library. And for an AI to go and query a huge library is very, very difficult. But how about, and it's so amazing because it's so elegant, he said, how about you build a Wikipedia page before the brain? So it's kind of like, When you go in a city, let's say you go into whatever, Barcelona or New York, you're not going to go every street up and down, up and down, up and down looking for restaurants. You're going to open Google Maps. And Google Maps is an intelligence layer. So it's going to show you above all the streets, hey, here are all the restaurants. And based on this, you now can do a much quicker search of the restaurants, what you like, where you like to eat, etc. And this Wikipedia page that we're talking about your company, that is that intelligence layer. Because overnight the embeddings are being updated, the vector database is updated, but overnight the brain learns and it's bi-directional. Because we know when we get a bad answer from AI, we cannot go back to chat GPT and say, well, what you wrote here is wrong, you you need to update your memory, it doesn't work. We cannot do it with copilot either. But when you have a second brain, you can make it bi-directional, which means you can go back to the brain and say, what you wrote there about this customer is actually wrong, actually this and this is correct. And in this one moment, it's going to go and update the Wikipedia page and then overnight the embeddings, which means your brain is learning. So it's not what you have in your company GPT, which is static because A plain vector database is a bit like a safe in a bank. What happens in a safe in a bank? Well, you walk in, you deposit something and you leave. Or you walk in and you take something out and you leave. You cannot walk in a bank safe 110,000 times a day. It doesn't work. And you need AI agents that are being able to query it. So, safes, what companies have been building as a company rag, Don't scale. Raw rags don't work at scale. And that's why your copilot feels dumb. That's why when you kind of ask a question to Microsoft Copilot, it's not Microsoft Copilot, it's any AI in the world, even Cloud Code or ChatGPT or whatever. When you ask a question, you get something which is a bit random, right? Because your company GPT right now is static. But the man who cracked it, Andre Capati, by the way, for those who don't know, Andre Capati was ⁓ the ex-head of AI at Tesla. And now he's kind of like an AI Pope. And he explained two things. The two fixes are that Wikipedia layer in front of the rag. So your AI comes in, it doesn't go and read all the million files which are in the vector database. But what it does is it goes and reads the Wikipedia. Even for our company, you will not believe it. The Wikipedia is super short. It's like 110 pages or 120 pages. of like, not even every single customer has its Wikipedia. It's like there's a cluster of customer. Every single employee has one. Every larger project has one. But basically it's like that Google Maps. It's like the AI comes in, quickly reads the Google Maps and then knows where to get it. And the beauty is obviously, you know, The AI writes the Wikipedia, you don't need to write the Wikipedia. And the second fix that he brought forward was this bi-directionality that corrections can flow back into the Wikipedia and the embeddings. So your brain gets smarter every time somebody gives it feedback. Now, how do you start to build such a prototype and how can you get value out of such a thing? Okay, now you can start a POC on SuperBase. Now, actually, you know, about a month or month and a half ago, Nate Jones, so Nate Jones is somebody that I admire a lot. I consider him a mentor. And he had that concept and he still has it of the open brain. You can go on GitHub, you can type GitHub open brain, and he will kind of explain this. He will kind of explain, look, go on SuperBase, it costs you $20, $30 a month, and you can quickly get a second brain that you can query, right? Now, type of customers we work with, they will kind of say, yeah, it's very nice, but you know, I want my data to be in Switzerland, I want my data to be in Germany. Okay, fine, so build it on Azure. As I said, one of the ones that we built is on Azure, and the beauty about building a second brain on Azure, is that you have the Microsoft phone line to tap to. And that's the Graph API, which means you can say, read all the emails, read all the calendars, read all the Teams messages, read everything that's on SharePoint. And 99 % of all enterprises we work with are on Azure. And the Wikipedia pages or that vector databases can sit on your tenant. Actually, Microsoft has a protocol to write rags, graph rags and vector databases. In this case, your data never ever leaves your cloud. And like I said before with Gemma 4, if you're really paranoid, A, you can run that on a local machine with 64 gigabytes of RAM. And you would only use the graph API to go and retrieve all the emails so that you can train your AI or your second brain with the internet switched off, like literally unplugged on your desk. And now let's look at the monthly cost of building such a thing. Guys, you're gonna fall off your chair and girls, if you take Glean, Glean you will pay roughly 10 to $30 per user per month, right? So for my 10 people employees and team, it will be around $300. And it's a stupid brain. It doesn't have any custom connectors. It doesn't have QuickBooks. It doesn't have Artio. It doesn't have SAP. It doesn't have all of these things. And fine tuning a model three years ago would have cost you a million dollars. Easy. Now two years ago, maybe a hundred thousand dollars. Today we're at a point where fine tuning a model essentially is free. Like that's the pace. your IT department does not have data security anymore as an excuse and as a constraint. And secondly, they don't have the costs as an excuse or as a constraint. Why? Because now you can essentially do that for free. Okay, Malcolm, nice. Now you have a second brain. What the hell do I do with it? Well, let me give you a very, three very concrete examples, threes again, consultants. Okay, the first one is build a second brain and walk in that next project management meeting or that next customer meeting. So you will walk inside or I walked inside and I asked the brain, I'm like, look, I'm about to meet this customer. What do I need to know? And again, the answers don't come like chat GPT in a minute. It takes 20, 30 minutes because you have different MCP calls, right? And one MCP call, you know, one update is from the Wikipedia. It reads the Wikipedia pages. Then the second one is the vector database. Then the third one is the contacts. Then the fourth one is the timeline. So currently ours has roughly eight different layers when it goes and gets an answer, right? But it's wild. when you get an answer in front of the customer that cites every email you had with them, every meeting, every recorded call that you had with them, and obviously everything you invoiced them over the last three years. So when I can show them such a thing, I'm not guessing, like the quality, the quality level of our discussion is immediately like super boosted. Like it's not like, ⁓ I think the next AI use case should be A. It's like, no, second brain, go and read or understand the 2,500 AI use ⁓ case that we have protocoled over the last two years over all the customers that we have. Find out where the pain point is of the customer and try to understand what is the next step. Now, McKinsey does that too, but they will take three months. It's cost you half a million dollars. and you get a PowerPoint. And today you can do it too. In 30 minutes on the toilet or in the train or in front of your customer. Why do I say on the toilet? Because you can build a connector that on your phone over dispatch over cloud code, you can query this before you meet the customer. And dispatch connects over the MCP server to your second brain, builds you an HTML dashboard as analysis and sends it to you on your email. Okay, let's look at another use case, not company internal, but company external. Now, I showed this a bit to a buddy of mine, he used to be like a serial CFO in different type of companies, et cetera. And he said, Malcolm, you know that one of our companies just hired a new CEO. And you know how it is when a new CEO, it doesn't have to be a CEO, it can be. anybody who starts in a new position. You need 6 to 12 months to even understand what the hell is going on here. And obviously he's saying this could be interesting because you know the new CEO obviously we can't ingest everybody's emails because it's a consent nightmare. You know maybe it's a company that is in multiple countries etc. Where the hell do we start? So my answer was you know what how about you just start with the outgoing CEO's inbox. Hey man, just start with that. Like, because the guy's leaving anyway. So take the inbox of the CEO. Why? Because an inbox of a CEO is worth gold. What's inside? Presentations, forecasts, problems. Hey, when somebody emails the CEO, it's because something's broken. Like a CEO's position is basically a problem identification. and solution machine. And I said to him, you know what, just build a second brain just from one inbox. And doing this, like no joke, is two days of work. Which means the new CEO walks into any department, any country, any production site and asks, hey, what are the 10 biggest problems we have here? Like what questions do I need to ask? to make people sweat. What is the probability that this business unit will get in trouble in the next 3, 6 and 12 months? And that you can only delegate to a second brain because you do not have enough analysts, enough head of strategies and people are not open enough to tell you that. But just by querying the second brain off basically the inbox of the CEO is going to open up all these inefficiencies and you're absolutely going to kill it. Guys, six months of onboarding compressed in a laptop. Okay, now then we looked at the second brain, we built it, now we built as I said two or three and now kind of Marsha and Marsha is the co-founder of this company. said, you know Marsha is, she always says she's my first brain because you know Malcolm has a small brain so he needs a first brain. Obviously all of your wives are your first brains but she said, okay it's nice Malcolm, now we've built the second brain of the company but I don't care. the second brain must be able to do things, which means why the hell are you still answering emails? Like this thing knows everything about our firm. Why do we have an analyst who's still reading contracts that come from customers? Because when we used to throw this in code code or we threw it in codex or whatever AI, it's not bad. The analysis is not bad, but it's out of context. But now when you ask the second brain to do this, The answer goes over 20 minutes, 30 minutes, but then it starts really having a lot of context. And she's right. She said, look here, a brain that only answers questions is 50 % of the value. But now a brain that does things, draft the emails, read the contracts, flag the risks, write the proposals, et cetera, that's 100%. And when it's a democratically elected second brain that has knowledge about everything and you give that to every employee, my God, you're on steroids. Like I can tell you the phase we're in right now in this company, we're on steroids because we have the second brain and slowly we're starting to give it tools. For those who are a more technical, basically give it skills that you've built in the company and let the skill act on the second brain through an MCP server before It does anything. Okay, now how do I have such a strong confidence that we're moving in the right direction? Because of SoftBank. SoftBank? Well, we know SoftBank is one of these biggest investment funds on earth. And kind of like what SoftBank normally invests in is, you know, like two or three or four years before the rest of the world. Like they're kind of a bit like Nvidia. You kind of look a bit what Nvidia is doing to understand in which direction the AI tech stack will move and SoftBank is the same. kind of look what SoftBank is doing and you say, okay, this is probably the direction in which things are going to move over the next few years. And SoftBank just invested like billions in a company called Skilled. Now, what does Skilled do? Skilled builds second brains for robots and not just them. mean, everybody's tuned in. NVIDIA is ABB Robotics, Universal Robotics everybody. Now, why do robots need second brains? Because they need context. So, a humanoid robot in a hospital needs to work very, very different than a humanoid robot in a nursing home and a humanoid robot in a factory. So, the unitry that companies are buying needs to act differently. So, it's the same hardware, everybody the same hardware. different contexts. And that context is the so-called brain. So robots need to understand a bit, okay, I'm now running around in hospital and maybe there are elderly patients that they need to watch out for. But at the same time, there could be like five doctors running past that robot because they're running into emergency care, for example, or into a surgery. So robots have the same problem as your knowledge workers. You dump 10 documents into chart GPT and it starts talking garbage and humans stall at 10 documents also. Give 10 documents to any human being in your company and say, can you work through this in the next half an hour? Obviously they stall if they don't use AI. But robots also stall at 10 data points. So both need a second brain to scale. And this is the uncomfortable truth. Robots will have a second brain before most of your companies and most of your employees will be given a second brain. Like isn't that crazy? We're behind stupid robots like mechanical robots. And back to this 70,000, you know, at the beginning I talked a bit about football. I mean, I don't... To be honest, I don't know anything about football. I know that Messi used to play and Ronaldo, maybe they still do, I don't know. But now, I kind of asked AI before this episode, said, okay, how much percentage of humanity is using code code today? And it's 0.35 % according to AI. Then I said, okay, how much percentage of humanity right now is building a graph rag? because ontology is kind of needed for a good second brain, plus a second brain. So, Claude code, graph rag and second brain. And then it divided that number by 100. Now, when you go into the Allianz Arena in Munich, there are 70,000 seats. You can watch Bayern Munich and there are 70,000 people around you. It's a big stadium where whoever's been inside, right? At these numbers, currently where we stand, if you are building your second brain, then congratulations. Because you are the only person in a stadium of 70,000 people who's understood that this is important. And now you could say, okay, knowledge in the AI world is what is three months, six months, okay? So in six months, anything I told you in this podcast is irrelevant, but... In these next six months, you are brutally ahead of each and every single competitor if you build a second brain of your company. Okay, so where do you start on Monday? Well, probably start light. Probably don't ingest the entire company because first of all, if you do it wrongly on Azure, it's gonna cost you an absolute fortune, okay? But start with that project folder or start with one email inbox, et cetera. You know that project that's been running for 10 years? That's really, really, really good because the project folder or stuff that's on SharePoint, there's no employee emails. You don't need any meetings. It's legally clean because You already have that shared company knowledge that you've been using anyway, but your people have been using it by actually going on SharePoint and clicking themselves through. Probably a good place, like Nate also explained it, is to build it on SuperBase. Why? Because SuperBase costs you $20, $30 a month. know, HNSW index, it's cloud hosted, it's infinite capacity. It doesn't matter if your SharePoint has. five gigabytes or 20 terabytes. But do the embeddings yourself. Don't pay for Azure for this because it costs a hell lot of money. know, download yourself a local embedding model and then it might embed instead of two days, it might embed three weeks, but you don't care. Then go and connect this second brain with an MCP server and don't build 20 tools, just build three. Like the first one could be ask the Wikipedia, second one is search the brain. And the third one is get the contact of the person who's also involved in this. And any AI can call this thing once you have an MCP server. And the next one is build that Rolls Royce of knowledge, the Wikipedia layer. Kapati taught us this. This is non-negotiable for you because without this, your answers are going to be absolute garbage. Now you have the second brain. Who do you start giving it to? Give it to the skeptics. Give it to that top salesperson in the company or the super overly confident manager. Because I can tell you they will be shit scared of that brain. Because they know or the second brain knows if these guys can keep their promise. The second brain knows what they've been saying over the last three years. And It's not like, oh, okay, you know, maybe they were right, maybe they were wrong. No, it's going to quantify how much BS they've been saying lately. So, you know, don't use it as a weapon against them. Try to explain to them that this is a transparency feature. This will have everybody in the company be more aligned to corporate strategy, be more aligned to make sure we don't make the same mistakes like last time. Actually, I'm going to do a workshop with an oil and gas company. in about 10 days I think and there I will try to explain them that because in oil and gas when you make a mistake I mean these mistakes can cost you a million dollars a day. So companies in oil and gas will be very very very interested in building a second brain. And as I said Damian explained it in one of our team's calls. You know it's good to have a second brain and like I said you know my wife is my first brain but I claim that Every CEO in a company has no idea what's happening in his company or in her company. Because the moment it gets to the CEO, it's been filtered by 20 people. So a CEO sits so high up that these signals are very, very much diluted. Well, how about building a second brain and taking control of these signals? Because now AI can have a second brain. And like I was teasing a bit, a second and a third and a fourth brain, but unlike human brains, the AI second brain scales. And that's what every CEO needs. So yeah, everybody, I'm Malcolm Vechota. This is the AI Cookbook Show. As I said, ⁓ our episodes are in English and in German. We try to bring what we do in our company. We try to explain this. We try to also explain what we see with customers. Episode 117, live from Bregenz. So, most wonderful greetings from Austria. All the best and take care for now. Malcolm out.