Sales

How to build a second brain that makes AI work for you

AI knows nothing about your customers, your process or your product, and it will not learn them by working with you. A second brain is how you give it what a good colleague takes years to build up.

Paper collage: an engraved brain joined by threads to small paper notes, a hand adding one more, and a torn paper strip with the word CONTEXT

AI is amazing, and very few people get much out of it. I talk to a lot of salespeople, and when the subject comes up nearly all of them say the same thing: yes, it's great, I use it to write an email or two. Meanwhile, a lot of what AI makes for me now just goes out, presentations and emails included. I tweak a line or add something it could not have known, and that is it.

The difference is context.

When you hire someone, they know nothing about your customers on the first day. After six months, or a few years, they have made the mistakes, sat in the meetings and worked out what your buyers really mean. All of that lives in their head. AI starts where that new hire starts. It knows almost everything except the four things that actually matter: your product, your buyer, your process, and what happened in the two thousand meetings you have sat in.

Daniel Nilsson on stage at UNLEASH World in Paris, under a slide reading: AI can't transform your workforce if it doesn't understand your people
Making the same point on stage at UNLEASH World in Paris, October 2026.

And it stays there, because it does not learn from working with you. A new chat starts close to zero. Some tools keep a short memory of your preferences, and on many work accounts your conversations are not used to train the model at all, which is what you want for privacy. Microsoft and Anthropic both say so. If you want AI to understand what a good colleague understands, you have to build that context yourself, along with the processes that use it. For me, that is the second brain.

This piece is about building one for yourself: your customers, your process, your way of working. A brain for a whole company is a bigger job with more to consider, and I come to that further down. Start with yours.

I learned this the slow way. Over the past year I worked with AI to document how my companies operate. Ideal customer profiles, pricing rules, how we keep the CRM, how we run a workshop, how I write an email. It came to a hundred and six documents. I do not regret any of it, because you cannot organise knowledge you have never put into words, and most companies have never put theirs into words at all. That work has to happen.

But it is the first half of the job, and I spent a long time believing it was the whole thing. For months I still corrected or rewrote roughly three quarters of what AI handed back to me, and I assumed that number was a limit of the technology.

It was a context problem. The knowledge existed, but AI could not get to the piece a task needed at the moment it needed it. Once it could, the number fell. Today I rarely rewrite anything. I tweak it, or I add what it did not know.

The problem was named in 1945

In July 1945, Vannevar Bush published an essay in The Atlantic called "As We May Think". He had just spent the war coordinating American scientific research, and he was worried about a specific failure: the record of human knowledge was growing faster than anyone's ability to find anything in it. Specialists, he wrote, were drowning in findings they could not consult.

His proposal was a desk-sized machine he called the memex, which would store a person's books, records and correspondence and let them build associative trails between them. The important part is why he designed it that way. Bush argued that indexes and categories were the wrong model because the human mind works by association, jumping from one thing to the next along links. The machine should copy that while adding what human memory lacks, which is permanence. He even imagined a user photographing a whole trail and passing it to a friend to insert into their own memex.

Storage stopped being the hard part decades ago. Finding the right piece at the right moment still is.

The same idea shows up in Niklas Luhmann's card index, and later in the term most people know it by. Tiago Forte popularised "building a second brain" in his course and then his 2022 book of the same name. The phrase has since been flattened into "somewhere to keep your notes", which is a shame, because the original point was finding things again.

Timeline: the memex in 1945, Niklas Luhmann's card index from the early 1950s to 1997, Building a Second Brain in 2022, and a second brain for AI now

Why more knowledge makes the output worse

When you give AI a large document of instructions, it reads all of it, and everything in that document competes for the model's attention at the same time. If you have written a thorough twenty-thousand-word guide to selling into manufacturing, and you ask for a follow-up email to one manufacturing prospect, the model averages across every buyer you have ever described.

That average is what people mean when they call AI output generic.

You cannot fix that by writing less. B2B selling is complicated, and compressing it onto a page throws away the part that made it worth having. What has to change is what loads, and when.

Separate the front door from the house

In an earlier piece on AI in sales I argued for keeping a skill, the instruction file AI loads before it starts a task, tight. That holds for the part that loads every time. It took me a year to see the rest of it: there are two different kinds of thing in your documentation, and they want opposite treatment.

The first is instruction: how to approach a task, what good looks like, what never to do, where to go for the detail. This is short, it is stable, it changes rarely, and it should load every single time.

The second is knowledge: the substance itself. Every objection you have handled and what worked, every call transcript, your real pricing, the emails that sound like you. This is enormous, it is specific, and it should load almost never, in slivers, only when a task calls for it.

The mistake almost everyone makes, and the one I made at scale, is writing the second kind into the first kind. You end up with an instruction file that is really an encyclopedia, and it opens itself in full every time anyone mentions the topic.

The skill is a short front door that loads every time; behind it is a house of knowledge, and a task opens only the rooms it needs, here objections and customer profiles

The best thing in my own setup is the skill that teaches AI to write in my voice. An earlier version described my voice in detail across thousands of words and produced writing that read as machine-written, for the obvious reason that a description of a voice cannot produce a voice. The current version is a short front door that says which situation this is, and behind it sits a file of forty-five of my actual sent emails, sorted by situation. It reads the handful that match and writes. It was the first thing I stopped correcting.

What the second brain is actually for

A second brain, in this sense, is a store of what you and your company know, structured so AI retrieves the specific part a task needs. Luhmann is the proof that this works, and the numbers are worth knowing. Between the early 1950s and 1997 he built a card index of roughly ninety thousand slips. He published dozens of books and hundreds of articles from it.

The cards mattered less than the links between them. Each card carried a fixed number and pointed at others, so that pulling one card pulled a trail of related thinking he had often forgotten he had. He described the box as a conversation partner. What made it useful was that he never had to read the whole thing.

That is what you are building. A store with enough structure that a question about a compliance buyer in manufacturing returns the three calls and two objections that bear on it, instead of the whole category.

A task, a follow-up email to a compliance buyer in manufacturing, pulls a trail of three calls and two objections from the second brain; the rest stay on the shelf

None of this depends on one product. It is a way of storing the context AI works from, which people now call context engineering, and it applies whether you use Copilot, ChatGPT, Gemini or Claude. Mine runs in Claude Code on a Mac because that is where I work. Further down I show what it looks like in the other tools.

Yours first, then the company's

Everything so far works for one person, and that is where I would start. A company brain has more to consider: who may see what, who approves what goes in, and who keeps it true. If you take it that far, these are the decisions to get right.

Two decisions come first, because they are easy to get wrong and expensive to reverse.

Build two. I run a personal brain on my own machine, and we run a company brain for Up Strategy Lab. They are separate on purpose. The personal one holds half-formed thinking, notes from books, things I am still working out, and it can be messy because only I read it. The company one has to be trusted by other people, which means it is stricter about sources and maintenance. Mixing them produces a store that is too scruffy to rely on and too formal to think in. What connects them is a deliberate step, not a sync: once a week, at most five items from my week's work are proposed for the company brain, and nothing moves until I have approved it.

Decide what the model is allowed to do. This is the rule that makes the whole thing hold together, and it is worth writing on the wall:

Raw inputs are immutable. The compiled wiki is the source of truth. The model is a compiler, not an author.
The model is a compiler, not an author: raw inputs go through the compiler into a wiki that you and your AI use, and the wiki can always be rebuilt from the raw inputs

Three things follow. Every compiled page carries a reference back to the raw file it came from, so if a page says something surprising you can walk back to the source in one step. The model is forbidden from adding general knowledge it happens to know, so the wiki represents only what you have actually read. And because raw files are never edited, the wiki is a derived artefact you can rebuild from scratch when you improve your prompts or change your format.

Most people break that third one by editing raw files in place, and then a change of format means starting again.

Where to build it, and how to start

Whatever you build it in, the principle is the same: the AI pulls only the few entries a task needs.

Starting today: a single folder of markdown files, one topic per file, each with a short summary at the top and links to related files. Point your AI tool at the folder. It is free, and it works.

Structure without engineering: Obsidian. Its graph view is the closest cheap thing to what Bush described, it stores plain files you can move later, and Claude Code, pointed at the folder, does the compiling on a daily schedule.

Available to every session and every colleague: a database with a small server in front of it that AI tools query directly. This is what the company brain runs on. It is more work and you should not start here.

Three steps up: a folder of markdown files, then Obsidian and Claude Code, which the free kit sets up, then a database and a small server for every colleague

The material to feed the store is already sitting in your business. Start with transcribed sales calls and won and lost deals, because they hold the real objections and the language customers use. Then CRM notes, support conversations, and the documents you have already written.

A prompt that works

You are helping me build a knowledge store my AI tools will query later. Read the attached material. Produce a set of separate knowledge items rather than one summary. For each item: a title of two to four words, a one-sentence summary of what it covers and when it would be useful, the substance itself with the specifics kept in, and links to any other items it relates to using double brackets. Split anything that covers two subjects into two items. Keep real numbers, names and quotes exactly as they appear. Reference the source file each claim came from. Do not add anything that is not in the source, and do not generalise. Add the date of the source to each item.

Run it over one quarter of call transcripts before you attempt everything. You will learn more from the first twenty items than from any amount of planning.

One practical note on cost. Compiling is volume work, so run it on a cheaper model and keep your best model for asking questions of the finished store. The judgement calls that matter most stay with a person: checking that a reference page is still true, and deciding what moves into the company brain. Getting that split right is the difference between a system you run every day and one you run once.

Building it in Copilot, ChatGPT, Gemini or Claude

I have used ChatGPT, and we have Gemini in the company, but Claude Code is the one I chose and the one I run mine in, so that is where my experience is. What follows comes from each vendor's own documentation, checked in November 2026. Limits change, so check them before you rely on them.

The method is the same everywhere. Build one assistant per job, with a short instruction and only the files that job needs. A follow-up writer gets your customer profile, your objections and twenty emails that sound like you, and it never sees the company handbook. That is the front door and the house from earlier, inside a tool your IT department has already approved.

Microsoft 365 Copilot. Create an agent in Agent Builder: write the instructions, then add the knowledge. One agent can use up to 100 SharePoint files and 50 OneDrive files, and it respects the permissions already on them, so the files stay where your team works. Its list of supported file types has Word, PDF, PowerPoint, Excel and plain text but no Markdown, so use Word or text files here. Grounding an agent in SharePoint needs a Microsoft 365 Copilot licence or pay-as-you-go billing.

Microsoft's own advice is to split long documents, because Copilot can give less attention to the middle of a long file. For your own work, a Copilot Notebook collects references to your files: up to 300 count with Microsoft Copilot, and 50 on Copilot Chat.

ChatGPT. A project keeps related chats, files and instructions together. How many files you can add depends on your plan. Make one project per job, so that each one only ever sees what that job needs.

Gemini. A Gem has instructions and a Knowledge section, and a file added from Google Drive stays current when you edit it. Google says Gems on personal accounts will move to skills from November 2026. NotebookLM suits a larger reference library, with 50 sources per notebook on the standard plan and more on paid plans.

Claude without Claude Code. A project holds knowledge and instructions, and free accounts get up to five. On paid plans it switches to retrieval on its own as the knowledge grows. Skills come closest to what I describe here, because Claude loads only the information a task needs. Neither needs Claude Code.

The part you will do by hand is the compiling. Run the prompt above over a quarter of call transcripts, check the items and save them as files. Then keep them true: a quarter of an hour a week to take what you corrected in AI's text and put it back into the files. If a team shares them, make that one person's job.

If you want to read further, Anthropic's Effective context engineering for AI agents explains why the smallest useful set of information works best. Ana Mourão's Context engineering is the real AI advantage in marketing argues that marketers should lead this work themselves.

How mine runs

Nothing about it depends on me remembering. Every conversation I have with Claude Code is captured as it happens. At 09:15 each morning, a compile turns the previous day's conversations, and anything I saved, into pages: one for each account, person, competitor and idea, with every claim linked to the file it came from. Every version is kept, so a bad compile can be undone.

Before a call, one command prints what we know about the company, and it leads with what nobody has asked yet. An account page never leaves a field blank. Anything we do not know says "UNKNOWN: ask", because a blank reads as nothing to say, while "UNKNOWN: ask" reads as a question nobody asked. Those fields become the agenda for the call.

The brain is one layer of what AI reads when I work with it. First comes a short file of standing rules: what never to do, how I write, where the truth lives for prices and customers. Then about 240 short notes on corrections I have made, decisions and how each project stands, most of them with the reason attached. That is the closest AI gets to learning from experience. It does not remember the correction, but it reads it next time. Then come the instructions for the task at hand, and the brain for what happened with this customer.

A brief from the Second Brain kit for an example company, led by the questions nobody has asked yet
Example data. The brief leads with what nobody has asked yet.

I started mine on 23 August. Six weeks later it held 691 pages, 79 of them accounts, compiled from 124 conversations and 249 other sources.

By the start of November, ten weeks in, it held more than 900 pages, 116 of them accounts.

We have packaged this setup as a free kit, so you can run the same thing: an installer, the compiler instructions, the page templates, the health checks and the daily compile, with step-by-step instructions written for people who have never used Claude Code. It takes about fifteen minutes on a Mac. Get the Second Brain kit.

The kit needs a Mac and a paid Claude plan. If you are on Windows, or your company runs on Copilot, ChatGPT or Gemini, the section above gets you the same method without it.

What this does not fix

A knowledge store nobody maintains becomes a confident liar. Old facts do not announce themselves as old, and an AI quoting a price you retired last year is a specific and expensive way to lose credibility in front of a customer.

Discipline fails at this, reliably, so the fix has to be structural. Decide where a fact lives according to how fast it changes.

Anything that moves weekly never goes in the brain at all. Deal stages, invoice status, who owns an account, what a customer is currently paying. That data belongs in the CRM and the billing system, which are already the source of truth for it, and the AI should query them live at the moment it needs an answer. The brain holds the pointer, and the CRM holds the value. A copied number is out of date the moment you copy it.

Anything that moves slowly does belong in the brain, and gets maintained deliberately: product truth including current limitations, positioning, the objections that were answered well, how you run a discovery call, what your competitors actually do. This material is worth curating because it changes over quarters rather than days.

A third kind needs no maintenance at all, and it is the most valuable: dated evidence. What a buyer said on a call, why a deal died, which objection came up and what answer was tried. A record of what happened on a date stays true about that date, so it can be stored freely, and it is usually the only copy anywhere.

Three kinds of fact: live state never goes in the brain, slow-changing reference is kept dated and maintained, and dated evidence is stored freely

Everything carries a date, without exception, so that anything can be aged out later. And the traceability rule earns its keep here too. When a page looks wrong, you want to find the source behind it in one step rather than arguing with a paragraph nobody can account for. If nobody can check where a claim came from, people stop trusting the store, and then they stop using it.

And none of this gives you judgement. I can tell when a draft is wrong because someone spent years telling me when mine were. That is why I still catch what is wrong or missing. A second brain lowers the correction rate. It does not replace the person doing the correcting, and if the people coming up behind you never get corrected by a human, retrieval will not teach them judgement.

Most of the visibility problems I see in B2B companies come down to something like this: knowledge the company already has and cannot find. I have twenty years of positioning work behind me, a Red Dot for MuchSkills, and 500,000 readers on the guides on this site. None of that stopped me building a library with no way to search it. Write everything down. Then build the thing that finds it.

Frequently asked questions

What is a second brain in the context of AI?

A structured, searchable store of what you and your organisation know, built so an AI tool can retrieve the specific part a task needs rather than reading everything. The idea is usually traced to Vannevar Bush's 1945 memex and was popularised as "building a second brain" by Tiago Forte.

How do you prepare a knowledge base for AI?

Split it into one topic per file, put a one-sentence summary at the top of each, link related files to each other, and give every entry a date and a reference to its source. Keep the original source files unedited, so you can rebuild the whole store when your format improves.

Should a company have one second brain or two?

Two. A personal one for half-formed thinking, book notes and things you are still working out, where mess is fine because only you read it. And an organisational one that other people rely on, which needs stricter sourcing and real maintenance. Combining them weakens both.

What should never go into a company knowledge store?

Anything that changes weekly: deal stages, invoice status, who owns an account, what a customer is currently paying. That data belongs in the CRM and the billing system, and the AI should query them live. The store holds a pointer to where the current value lives.

What should I put into a second brain first?

Transcribed sales calls and your won and lost deals, because they contain the specifics no one ever writes down: the real objections, the language customers actually use, and what was said just before someone decided.

Do I need to be technical to set one up?

No. Our free Second Brain kit sets up a personal second brain on a Mac in about fifteen minutes, with step-by-step instructions written for people new to Claude Code. You need a Claude plan that includes Claude Code.

Can I build a second brain in Microsoft Copilot or ChatGPT?

Yes. Build one assistant per job: an agent with SharePoint files in Microsoft 365 Copilot, a project in ChatGPT or Claude, or a Gem in Gemini. Give it a short instruction and only the files that job needs. The free kit on this page is for Claude Code on a Mac, but the method works in any of these tools.

Does AI learn from working with me?

Not the way a colleague does. A new chat starts close to zero, some tools keep a short memory of your preferences, and on many work accounts your conversations are not used to train the model at all. If you want AI to know your customers and your process, you have to give it that context yourself, and write your corrections down where it will read them next time.

Why is my AI output so generic?

Usually because AI has either none of your context or all of it at once. Give it a big document and everything competes for attention, so a follow-up to one buyer comes back as an average of every buyer you have described. Keep a short instruction that loads every time, and let AI pull only the few pieces of knowledge the task needs.

What is context engineering?

Deciding what information AI has in front of it when it does a task. In practice that means short instructions, a store of knowledge split into small files, and only the pieces the task needs loaded at that moment. A second brain is one way of doing it.

What is the difference between instructions and knowledge?

Instructions say how to do a task, what good looks like and where to find the detail. They are short and load every time. Knowledge is the substance: objections, call transcripts, real pricing, emails that sound like you. It is large, and AI should read only the pieces a task needs.

How do I stop AI repeating the same mistakes?

Write each correction down where AI will read it next time, in the instruction file or a short note with the reason attached. AI does not remember your correction by itself, so the learning has to live in your files.

Does the Second Brain kit work on Windows?

No. The kit needs a Mac and a Claude plan that includes Claude Code. On Windows, or if your company runs on Copilot, ChatGPT or Gemini, this article shows how to build the same method in those tools.

Should I build a second brain for myself or for my whole company?

Start with yourself. It is one person's customers, process and way of working, and mess does not matter because only you read it. A company version is a bigger job, because you also have to decide who may see what, who approves what goes in, and who keeps it true.

The Visibility Edge

Tested ideas on building better B2B companies, when there is something worth sending.

Similar posts

All articles →

Strategy

What people say versus what they buy: how to test demand

People are polite, and "I would definitely buy that" costs nothing. How to test your offer and your ideal customer properly, with the stories of Dropbox, Buffer and Warby Parker and what they mean for B2B.