РАСШИФРОВКА 🔗
Hannah: having worked in the industry for so long, but it still astonishes me how many versions of an ICP and bio personas are living within a company. And every salesperson has sort of an adjusted ICP. Marketing has a completely different understanding. And then leadership is a whole nother ballgame. So by building a context layer, you actually need to make decisions. Hi, welcome to another episode of Lead with Trust. In today's episode, we are going to have a look at what is a context layer, what's a company brain, and how are they different, and how are they different from all the other names and labels we give this thing. That we might describe as a context layer or a company brain. So bear with me today. We're going to take it all apart and we're going to put it all back together. And by the end of the episode, you will know exactly what is a context layer. Which part of this do you need in which situation? When is a company brain enough? And how do all the other terms fit into this world of context? All right. So let's get right into it. when I started writing my book, Scattered to Scaled. In pretty much I was starting to think about it in November, December of 2025, and I really started to earnestly put sort of together the concepts in January and then I wrote it in probably February, March. I started. It took me 5 months and by mid-August I sent it into my editor. In the very beginning, when I really started to think about this concept of what makes AI scalable, why are some companies stuck in this scattered place of you know employees trying to give their AI all the AI needs to know, but there's still AI slop coming back. And they're stuck in really low-value use cases and they can't figure out how to make the switch. To higher value use cases, right? And how to unlock sort of that full potential that AI can provide in a go-to-market scenario or in other functions of the business. Because there's a very, very small, depending on who you ask, six to ten percent of companies who really figured out how to scale, how to use AI and get compounding business results. But what are they doing differently? So I kind of became obsessed with that idea of what is the thing that unlocks it. And it was pretty clear to me very quickly that context, right? The AI knowing exactly what who you are, what makes you different, and what sort of market do you play, who are you up against, what's your secret sauce? What would you never say? All these things. We never provide that to the AI. And that became really quickly an obsession of mine in a healthy way. That I really started writing the second book from scatter or scattered to scaled. Right? So, but my point of the story is that when I started thinking about this, no one talked about it. Like barely anyone, the fringes were talking about it. And The fringes were really discussing this very hotly, but the mainstream was just bitching about AI slop and you know you just need to not be lazy and all that kind
Load full transcript
of stuff. But that's not the point. So I even got to the point mid-March-April where I called my mentor and I'm like, I need to walk you through this because I think I'm off my rocker here. No one else is talking about this. How is this possible? By the time I I sent my manuscript in for editing, the world had completely changed. All of a sudden within three months, context was almost becoming an overused term because now people caught on. Right? Now people were like, context is the thing that we're missing. But now that we're in November twenty twenty six, context and context layers and company brains and agent memory and all the other terms are almost like the Wild West. There's no defined definition that's accepted across the board. And when you say context layer, you can mean anything. If you say company brain, You could mean the same as context layer or something completely different. So this episode hopefully contributes to clearing up some of the confusion and will help you to understand what each term is, when do you need what, and how are they different. Alright, so before we go and dive into what is a context layer and the actual definitions, I just want to give you a few definitions that are out there in the market just that you understand where the nuance is and how broad that definition actually can be. So Redis is defining a context layer as the thing that decides what an agent knows at any moment. So they're really talking about run for runtime infrastructure, retrieval, memory, session state. Teleus says that a context layer is the architectural tier between enterprise data and AI agents. So they're really approaching this whole thing from a semantic layer plus the entity resolution and policy, right? Cube defines context layer as the governed business context between raw data. And AI agents. So they're here talking more about metrics and permissions and joins over the data warehouse. Atlan defines it as the thing that delivers governed meaning, lineage, and policy to agents at decision time. So again, it's mostly about data governance, and it is handed to agents as they're acting. Siemens with Rapid Miner defines context layers as the semantic layer that sits above your existing data and platforms. Again, it's basically an enterprise knowledge graph. FiveTren defines it as agents answer based on your actual business definitions, operational data, and analytics. So here we're talking about a shared layer that all of the AI tools can connect to. Zoom info defines it as what makes AI agents safe to run GTM at scale. Okay, verified data on accounts and concept contracts, and they're keeping it up to date, they're keeping it fresh. Mind Studio refers to a context layer as everything that sits between a raw user request and AI models. So that's very similar to. one above. So here we're talking again about memory retrieval and state and permissions. Unabyss talks about a context layer as something that gives something to know or gives AI something to know. So who is the wrap and it's basically served to every AI tool as the wrap uses it. And HubSpot says that growth context that's the GTM sort of lens on context that they coined in Unbound as the dynamic knowledge AI has around your business, your team, and your customer context. And with dynamic, they're really referring to it's maintained automatically. So but here we're referring really as to context or the context layer as one platform's context. scored for completeness. Okay? So as you can see, those are relatively broad in terms of what is a context layer. But when you look at them closely, you really see that they fall into two camps. The first camp is really more about enterprise data governance. And the second camp is GTM, like a GTM lens on things that gives AI what they need to, what it needs to know. In order to help you go to market. So we're going to focus here on the go-to-market lens because my content is specifically made for sort of mid-size, early growth, pre-scale B2B companies and not for enterprise. Both of those types of context layers, though, are valid and they're very critical and they're important. But an enterprise would start first usually on tackling context on a data governance level, whereas a mid-sized company would really start with looking at the GTM context first. So that's what we're mainly going to focus on here today. Okay, so now we looked at all the different variations of what a context layer is. Now let's have a look how I define a context layer. So, a context layer, in as I define it in my book, Scattered to Scaled, is the explicit, deliberate representation of what makes your company uniquely recognizable. So, what do you know? What do you stand for? What do you hold by? And all of that extracted, codified, and structured, implemented and amplified so both you, your people, and AI can accurately and consistently sound. behave and act like you. So essentially you're giving AI everything it needs to be trustworthy at scale. Okay, so you you're really defining who you are and being very explicit and deliberate about telling this to the AI so it can then act on your behalf and and decide on your behalf and sound like you accurately and consistently. Okay. Now I though have three kinds of context that I put together as a context layer. The first sort of is what we've always had knowledge bases. Especially if you're a tech company or software company, you always had documentation, but most of the companies in the B2B space, mid market. had knowledge bases, always had. So that's a really good starting point because it already documents and hopefully it's accurate and up to date. So this really gives you a good base of information around your products, how to use them, how they work, the features, feature releases, All these kinds of things. How to maybe handle billing. So that gives AI already a massive amount of information for you for it to act on. Now, in addition to a knowledge base,
there's something that is bubbling up right now that I'm really excited about. And I call that a context engine. A lot of people call it customer memory, or well, we get to that in a second, what what other people call this. But let's just call it context engine for right now. I'm not trying to if define a new term here, but it makes sense in a minute. And why I call it engine is because this solution is able to generate from our activities, from our GTM activities, mainly call recordings, meeting notes, your meetings in your calendar, your emails. All of the exhaust of what you do on a daily basis, like what is actually happening in your environment, and it has a look at those and it extracts the context. So, what actually has happened? Has this meeting happened? context on your customer in terms of what pain points did they bring up in a call? What was the solution you presented in the presentation? To them when you talk the next time. All of this context is now captured. and that's really important because a lot of the unstructured data, and it's about 80 to 90% of the business value, will used to be locked up in that unstructured data. And now for the first time, we are gaining access to that. We can mine it, we can create context from it, we can create knowledge graphs from it. And that is super exciting because that gives us a massive base of of knowledge and insights and context that we can now leverage and make useful. Okay? So here you could, as I said, look for pain points that your customer has, detect churn risk, upsell opportunities. You could have opportunities for testimonials. If someone says something nice, maybe you reach out automatically and say, Hey, you mentioned this and this. Would you mind giving me a testimonial? you could, I mean, there's there's so many different variations of what you can do. Now, I always refer to this as the candy. Everyone wants to do that because it's easy, it's awesome, it gives so much value so quickly. And it's auto-generated, right? You don't need to do anything. So that is awesome. And that's about 85% of the context that we're generating. Now, the problem of that is if that's the only thing we're doing, that we're kind of like a little ship on the ocean and we're drifting around and we're not steering towards anything because we're only driving. Almost with looking only in the in the review mirror. And while that's really helpful to have a review mirror when you're driving to see what happened, you definitely need to know what's ahead of you. So that brings me to the to this third layer, and that is your decided context. And that decided context will get over time thinner because the context engine will get smarter and will be able to capture decisions better. But there will be I think always a thin layer of what are we doing? how do we define as a leadership the value proposition? What are our goals? Where are we heading? Where's our North Star? And what are our values? Because if we just build context based on what our salespeople say every day, they might over time drift with our value proposition. They might have very different value propositions that they're telling their customers every day. And that's not a decided definition. So you can see how that layer is extremely critical because that is the layer that decides conflicts, and there will be conflicts in your context. That's also the layer that is most companies don't have it, it's the most ignored. That's also the layer that. I work with my clients on and we work through this in workshops and really make decisions and capture those decisions and put that context layer in place so that your AI has everything that it needs. Okay, so that's what the context layer is with its three components. But how does a company brain fit into this? Is that the same thing? So Company Brain actually has a quite interesting history, and I just want to share a really small tidbit here. So the Y Combinator in summer 2026, remember when I was talking before about how I thought I'm going crazy because no one was talking about it? Well, that changed in the summer because they put a request in for startups to say, Hey, we thought about this idea, and context is really important, and we want you guys to come up with a company brain as the new primitive for every single business. the company brain is basically the one share place that your AI agents read from and it's filled with whatever the vendor captures and gathers about your company. Does that sound familiar? That's what we just discussed, right, with the context engine. So you could almost say that the company brain is related to the context engine, but I'll show you in a minute how that all fits together. how does a company brain compare to a context layer? So a company brain really refers to the one shared place that your agents read from. And depending on the vendor that builds the company brain, it will include usually two of the three types of context that I have mentioned earlier. Now, before I go into that, I just want to mention the GTM brain here, real quick, because that is a term also thrown around a lot. when we're talking about company brains and A GTM brain basically refers to a specific type of company brain and it is solely focused on the sales, marketing, and service side of things in the company. So it doesn't look at legal, it doesn't look at finance, but really focuses only on the context, memory, and governance layer from a go-to-market perspective. Now, if you think about how a company brain And a context layer relate to each other, you can imagine it in sort of this kind of way. A context layer has those three types of context that I was mentioning before. The auto-generated, that's the context engine, the assembled, that's your knowledge base, and then that the decided, which is your canon. However, a company brain usually only touches two of them and maybe scrapes the third one. So company brain usually is having or
it includes the customer memory, so the auto-generated, and the knowledge base, so the assembled. And sometimes it allows you to add a few things that are decided. That are very context specific in what they want to fill the company brain with. Because a lot of vendors have different angles of the company brain depending on what they're speci specializing in. Now I just want to bring home one point that is really important. And I allured to it a little bit earlier. that the canon, the decided, the single version of truth, the most ignored type of context, I think is very critical because if you have a company brain and you only have the assembled and the auto-generated in it, and it may be a tiny slice of decided, you are going to have. probably an okay time for three to six months. It's probably really exciting and you get so much context you don't even know what to do with and you're excited. But after three to six months, you're starting to sort of feel there's conflict. Maybe you because the the context engine reads the emails that have passed, right? And now maybe you change your pricing. Or you tweaked your value proposition, or you're offering a different product, or maybe you retired a product. So changes have been made in the organization, but your context engine still works as a review mirror. It only works on what has happened before. Now, as there's new emails coming with pricing, or as you mentioned, things, a lot of context engines are really good in updating this. But there will be times when you have conflict. And your decided layer, your cannon is the thing that really decides on when there's a conflict. It also keeps you forward looking. It keeps you in a position where you are the ship, where you are the captain of the ship. You are steering it rather than having been pushed around with the wind blowing you wherever it wants to blow you. Right? So that's really to under really important to understand that. That canon part of the context layer is what you're missing if you are implementing the company brain. So I always urge people to have the company brain, but add the canon to complete their context layer. Now I want to go a little bit into some other related terms here because it's really important that we mention them. So anytime we talk about memory. You basically can assume that they're referring to that the vendor is referring to anything that has happened before, but it can't tell you what's decided. A context graph is basically a term for decisions and reasoning behind certain pieces of context or pieces of knowledge, and it's captured as it happens. It can show you a decision though, but it can't really make one. So while context graphs are enormously helpful, again, it does help to have some decisions that you make rather than just representing the decisions itself. Context engineering is just a term that refers to the practice of choosing what reaches a model in the moment it acts. So it's not necessarily the storage
or it's not the substrated lives on it's just a practice that draws on all those three grades as AI is interacting with it. So Second Brain is a really funny one, and I absolutely love this book. I've read it years ago, and it's all filled with notes and you know markings for me because I've been building my second brain for years and I used Notion for it, loved it, and really captured a lot of things that are happening. Now with AI, that's obviously much, much easier. Than before because you had to be quite diligent. But the reason why I'm saying is this is a second brain is a personal system for storing and connecting what a person knows. Right? That's Tiago Forte's term. So it's the same grades as a company brain, but it's scoped for one person. And the reason why I'm mentioning this here is. If you might have listened to my last episode, I had on Bob Ruffolo who was explaining how he built a second brain to help AI adoption in his company, and how he really wanted to download what's in his head and really structure it in a way that's AI accessible. But in the process, he found that he actually made founder dependency worse because he's the CEO of Impact. And he made it worse because he only built a version for himself. So if it's for one person, it's a second brain. If it's for a whole organization, it's a company brain or a context layer. Now persistent agent memory is what an AI agent knows between sessions. What it basically keep the knowledge it keeps between sessions. So facts about you, pastos, conversations, that's a relatively common term and it's so sort of a subset of the company brain. So this concludes the definitions of the terms that I most often hear CEO come to me with and ask me to explain and ask me to sort of separate them out for them. But what I want you to really take away from if you're a solopreneur, if you have a tiny team like let's say me and two assistants, you could build a company brain because you still tend to make every call yourself. But I've even as a solopreneur feel that my context layer is tremendously helpful because I went through the same process as I apply it to my clients. And even though it's probably a little overkill for me as a solopreneur with two assistants, it was incredibly helpful because now I'm I literally connect. them with my context layer and they have everything they need to go pitch podcast guests they can go off and build landing pages they can whatever it is they need to do because they have that knowledge and another reason why it's so helpful is because I for myself had to make the decisions and had to make it clear. I had to become clear on the things in order to write them down. And that is one of the biggest benefits that I see with my clients. Initially, they're like, We're having two-hour workshops or we're having four-hour workshops to go through decisions. And after the workshop, they're like, Wow, we should have done this years ago. I cannot believe how. useful that was because we had to get crystal clear on our process. We had to get crystal clear on the definitions and share the definitions between teams now. We can align our teams because we are now running on all those guardrails definitions. We as a company have decided what our value proposition is, what our ICP is. I cannot believe how well I can having worked in the industry for so long, but it still astonishes me how many versions of an ICP and bio personas are living within a company. And every salesperson has sort of an adjusted ICP. Marketing has a completely different understanding. And then leadership is a whole nother ballgame. So by building a context layer, you actually need to make decisions. And that's why the decided layer is so important as a component of your context layer, because otherwise the context engine will just live with these multiple definitions of the ICP. Yes, good tools will bring it up to you and will say, hey, there's a conflict. But in order to really have a very efficient machine and be able to scale and be trustworthy at scale and give your AI agents autonomy safely, you need to have that decided layer. Now, if you listen to this episode and you're thinking, huh, I would love to know what that looks like. I want to know more. The first thing that I would recommend is really that you get the book. you can get it anywhere books are sold, but you can also go to my website and actually get access to a free audiobook. And the audiobook has then a companion guide where you have readers questions and self-reflection guides and you it comes with a assessment like a scatter to scale scorecard but it also gives you all the graphics in it so you can really follow along. Okay, so that's the first thing. The second thing is I would love to invite you to a webinar I'm holding November eighteenth. And if you're listening after this it will be recorded and will be available on the website. But I'm going to walk you through How to set up an agent in a context engine that will actually help you listen out for those things that I was mentioning. So that you can mine your calls for testimonial moments, pain points, questions that your chat bot should be answering, that your customers repeatedly are answering, all those things. Okay. And I show you how. generally to do that but also practically what you need to do and I'm going to do that in day AI who I recently partnered with and I want to share the story with you how I set up an agent in four hours and I've never looked back. So I'm gonna show you a really quick walkthrough with the high level points that you need to know and I'm looking forward to sharing that with you. I hope this was useful. If it was please subscribe and love to have you leave a review that always helps other people discover the a podcast and I will see you next time.