Company Brain: the guide for leaders

“We need a Company Brain”: what sits behind the sentence

“We need a Company Brain.” In technology companies you hear it almost daily, and more often every month. It usually comes after a decision failed to land: the leadership team set the direction, and three weeks later teams, tickets and AI agents still work from the old version. This guide is for leaders who want to turn that sentence into a plan.

The finding behind it is measurable. In the AI Monitor 2026 we surveyed 64 companies in Switzerland and Germany on their AI readiness. The highest of 19 factors: openness to AI, 3.94 out of 5. The lowest: actually adapting existing structures and workflows, 2.47. Same sample, two questions. People are open. The way of working barely follows.¹

This gap is exactly where decisions disappear. And it widens once AI agents join the work: an agent working from old context repeats the old instruction faster than any person could carry it out. Automate bad input and you simply ship the mess faster.

What does this guide answer?

Simon and I have worked with context in a range of management roles: with decisions that don't land and with knowledge that walks out of the door with people. It kept catching up with us. That is why we are building Teklens. We produced the AI Monitor 2026 together with a master's thesis at the University of St.Gallen. This guide sums up the core messages of my keynote at the Entrepreneurs' Summit 2026, and what I believe follows from them.

It answers the questions leadership teams ask me most often: what information do the team and the AI agents get from a Company Brain? How do they access it, and who is accountable? Buy or build? How do individuals pass important knowledge up without creating chaos? And how does quality stay high?

My thesis: a Company Brain is not a software decision but a leadership decision. Who decides what holds in the company? Who approves changes? And how does that version reach people and AI agents at the same time? You can buy or build the software underneath. You cannot buy or build the answers to those three questions.

What is a Company Brain?

A Company Brain is a company's shared memory. It connects knowledge, decisions and meeting notes from the tools you already have, records what was decided and why, and gives people and AI agents the same approved version. What matters is that the knowledge is governed: every piece has an accountable owner, a date on which it was last confirmed and a label for how binding it is. Without that you have a search engine.

It matters just as much what a Company Brain is not. It is not another wiki everything moves into. It is not a search across all sources that also finds what is out of date. And it is not a chatbot on the shared drive. It sits on top of your existing tools, between the sources and everyone who works with them.

The discipline behind it is called enterprise context management: keeping a company's context current, checked and usable. Organisational memory is the goal, the single source of truth the rule.

Why does a company need a Company Brain when everyone already has AI?

Because everyone has AI, but the company does not. In the AI Monitor six in ten respondents also turn to private AI tools, even though 87 per cent get tools from their employer. Every team builds its own assistants with its own prompts and its own sources. A leadership decision may reach the people. It does not reach the agents.²

The result is the strategy execution gap. It rarely comes from resistance. It comes from documents, tickets and systems never hearing about the decision. An example: the leadership team decides “Strategy 2027: focus on the DACH mid-market, no US expansion”. Six weeks later the Jira epic “US launch Q2” is still active, the Confluence roadmap says “USA 2027”, and the sales assistant writes proposals from the US pitch deck.

The AI Monitor shows where the companies stand that have this under control. We grouped the 64 companies into three profiles by their readiness scores: 12 Leaders, 27 Followers, 25 Laggards. The biggest differences are not in the model. They are in the data foundation, in people and collaboration and in adaptability. These are readiness profiles, not success groups, and they prove no causality. But they describe the gap precisely.¹

Size is no protection: 20 of the 25 Laggards employ more than 250 people. More people means more heads holding knowledge and more places where a decision gets lost.¹

What information do the team and AI agents get from the Company Brain?

Seven kinds of knowledge belong in a Company Brain. Grouped, there are three:

For the first version one rule applies: a Company Brain that honestly covers three of these layers beats one that promises all seven and answers from a forgotten draft. For companies that make AI readiness a matter for the top, that means: start with what holds and with the conversations where it is made.

What people and AI agents need from it is the same. That is the core message of my keynote at the Entrepreneurs' Summit 2026. Imagine the leadership team decides on Monday: customer onboarding first. On Monday a new colleague and an AI agent have the same three questions.

Access. Where is the current version, and which sources may I use? On Monday both see only old tickets; the decision sits in the minutes, not in the system. The answer is a shared version that connects CRM, tickets and Monday's decision, with one person keeping it current. In the AI Monitor this is the data dimension: Leaders 4.11, Laggards 2.31.¹

Orientation. What takes priority, and which goal applies to the task? Sales promised the customer something, product has the roadmap, the agent works through its queue. The answer is a Company Brain that links the decision to the strategy and the work in progress, shows deviations and hands goal conflicts to people. This is the people and collaboration dimension: 4.53 versus 2.73.¹

Ways of working. How do we change the work, and which rules and approvals apply? The decision is new, the workflow stays old. The answer is a playbook: the concrete change to onboarding, an owner, approval by people, an update in the original system and a review after one week. This is the adaptability dimension: 4.13 versus 1.84, the sharpest separator between the profiles.¹

This mapping is my interpretation of the study, not a tested intervention. But it shows why a Company Brain is not an IT topic: the leadership team answers each of the three questions, not the technology. To check the dimensions for your own company, the Management Grader measures them in three minutes.

Where are the decisions made that belong in the Company Brain?

In conversations. Important decisions are rarely made in a document. They are made in the leadership meeting, in the customer call, on the video call and in the corridor, online and offline. And most of them are never written down. That is why connectors to files never find them.

Capture the conversations. A meeting companion that joins leadership and planning meetings, transcribes them and records decisions with owners changes more than any connector. For conversations without a video call the same rule applies by hand: the decision is added the same day in two sentences, with an owner. Exceptions for sensitive conversations are clearly defined, and everyone involved knows the meeting is being recorded.

Show the deviations. The decision alone is not enough. Jira, the codebase and the documents show where the work deviates from it: the epic that is still running, the pricing logic in the code that still implements the old rule, the roadmap that still names the old goal. Only comparing decision and work makes the gap visible, with source and location. That is why these systems belong in the Company Brain, not just the documents.

Decisions as a log. Every material decision gets a page: context, options, choice, rationale, date, owner, and what no longer applies since. That is the decision log, the layer with the highest value and the lowest coverage.

Old archives are read once, not at every question. Knowledge often sits in four years of a chat channel or in a thousand closed tickets. Let agents search the archive once in the background and propose candidates that a subject owner reviews. One rule of thumb decides what is queried live and what is curated: Jira, CRM and code are queried live; priorities, decisions and commitments are curated as a short, correct page. Nobody needs an agent that reads forty hours of transcript to answer “What are we working on this month?”.

How do people and AI agents access the Company Brain?

Where they already work. A Company Brain that requires learning a new tool does not get used. People see the approved version in their meetings, tickets and documents, and changes land where the work happens once they are approved. AI agents read through MCP, the open standard assistants and coding agents use to fetch context.

A good Company Brain does not reinvent who may see what. It takes over the permissions of the sources and adds three labels to every piece of knowledge: who may see it, how much weight it deserves (draft, checked, binding, superseded) and how carefully it must be handled. The three are independent. A page can be binding and strictly confidential, or visible to everyone and long superseded. Merge them into a single “access” setting and you get agents serving outdated policies to exactly the people who act on them.

For AI governance this means: every answer shows its source, every change goes through an approval, and the LLM data controls apply to every model that reads company knowledge.

Who owns the Company Brain?

One person. If you cannot name an accountable owner, you do not have a Company Brain project. You have a pilot that will stall. Architecture without accountability is a shared drive with a chatbot on top.

In practice there are three roles. The leadership team owns the scope, the budget and the rule on what belongs in the Company Brain. IT or the AI team owns the connections to the source systems, the permissions and operations. And every topic has a subject owner: for the price list, the product roadmap, customer commitments, HR policies. That person says “this no longer holds” and makes sure the correction applies.

The most common mistake in practice: the knowledge sits with the smart people, and they are not available right now. On holiday, in a customer meeting, in another time zone. Then the team waits, or it guesses. It slows things down even more when those people leave the company: the tribal knowledge leaves with them, and nobody knows any more why something was decided.

What works against it: every team builds its own knowledge hub in the tool it already uses, with an owner, ongoing upkeep instead of a one-off exercise and a clear rule on who decides when two pages contradict each other. Ten teams in parallel get further in a quarter than one central knowledge project in a year.

Second brain, personal knowledge management, personal AI agents: where is the difference?

A second brain is personal. The term comes from personal knowledge management: a system in note apps such as Obsidian, Notion or Apple Notes in which a person records ideas, decisions and lessons to take load off their own memory. With AI it becomes a curated context about your own work that your assistant can read, extend and keep current: a short file describing who I am and how I work, plus small topic pages and a rhythm of capturing, tidying and checking.

Since summer 2026 there are products for this that go far beyond notes: personal AI agents with their own memory, their own computer in the cloud and access to a person's apps.

These agents are impressive. And they do not answer the leadership team's question. They build one person's AI memory: what they decided, how they work, what they know. Three employees with three personal agents hold three versions of the same instruction, and at some point they contradict each other. Then there is shadow AI: an agent with access to email, calendar and company documents reads things it should never have seen, and the knowledge sits with the vendor, not with the company.

How do individuals pass knowledge up without creating chaos?

Through a bridge with a single rule: publish, never sync. Nothing from a personal memory becomes visible to others unless the person deliberately publishes it. What crosses the bridge is never the whole container but the knowledge: a short, cleaned-up note with the lesson about a customer's procurement process, not the note with a personal opinion about their head of purchasing.

To avoid chaos the bridge needs three things. An order of precedence: for your own work your own memory wins, personal beats team, team beats company; the Company Brain fills gaps, it does not override how someone works. A review: what is published lands as a proposal with the subject owner, not straight in the binding version. And reversibility: whoever shared something can withdraw it.

A study of four research groups that used a shared organisational memory for a month shows why this is hard: people preferred to ask questions privately, and contributions stayed away because turning personal experience into something generally useful is work. Nobody wants to look unpolished, and rewriting for others never feels urgent.⁴

The consequence for the leadership team: quietly harvesting personal memories teaches everyone to keep their real memory somewhere else. The bridge only works with consent and visibility. Start with one team, give people time to publish, and measure whether the agent can answer a new colleague's questions about what the team knows together. Good onboarding is the most honest test of a Company Brain.

How does quality stay high when AI agents rely on the Company Brain?

By treating it as production infrastructure. The moment an agent uses a page to act, to write a proposal or close a ticket, that page is production. The rules are simple: one owner per official page, advance notice before changes many agents depend on, mark things “superseded” instead of quietly overwriting them, a way to roll back, and two-person approval for legal, HR and security.

Then there is measurement. Most systems that feed knowledge to AI are never tested on whether they deliver the right material. The minimum: 20 real questions per topic with answers checked by the subject owner, which the Company Brain must pass before launch and regularly afterwards. Plus tests of whether it refuses what a person may not see. Three checks show a Company Brain is real: knowing correctly, refusing correctly, forgetting correctly.

For the leadership team this means: approval by people is not a brake but the mechanism that catches errors before they multiply. Nothing goes live without approval.

Buy or build?

Both. According to Menlo Ventures, 76 per cent of enterprise AI applications were bought rather than built in 2025; in 2024 it was still about half. Platforms are bought because they become productive faster. At the same time, coding agents make building cheaper, and many building blocks exist as open source.³

Whatever route you take, one thing stays with you: there is no vendor you can hire to decide what your company holds to be true. Topic owners, approvals, testing with real questions and upkeep run permanently, and nobody sells them to you. Software reduces the technical work. It never takes over responsibility for whether the knowledge is correct, current and owned.

This is exactly why we built Teklens: a Company Brain that sits on top of your existing tools, connects decisions and knowledge and writes approved changes back to where the work happens. A Company Brain is not installed, it is built in, together with the company: into your systems, with open-source components and with operations that must be secured. We are happy to help with that. Because it sits on top of the tools and writes back to Jira, Confluence, SharePoint, Notion, GitLab and GitHub, governance is on the table from day one. That is why it matters where it runs once things get serious: as Secured Cloud or self-hosted in your own environment.

That is the reason for the strong demand from companies that are already far along with AI: models get better, but everyone has the same access to them. A better model is no edge when your competitors get it the same day. The Company Brain can be the deciding factor for competitive advantage, because it is the part nobody can copy: your context. How the build works is on the product page.

How does the Company Brain connect Discover, Define, Build and Operate?

Good leadership connects knowledge, decisions and work. In the AI Leadership Lab four phases describe this leadership work, and the Company Brain sits in the middle: every phase draws on it and adds new knowledge.

Every decision builds on what your company already knows. So that your team and its AI agents can build on it for the next task, the Company Brain has its own role in every phase:

The proof: where Leaders score higher

The strongest proof comes last, and it is no sales pitch. Adaptability separates the readiness profiles in the AI Monitor most sharply: Leaders 4.13, Laggards 1.84, and the quartiles do not overlap. Behind it sit four factors: recognising change, deciding, restructuring, learning. The lowest single score in the whole study remains actually adapting workflows, at 2.47.¹

The companies out in front do not have the best model. They have a way of working in which a decision changes workflows. A Company Brain is the tool for that, for people and for AI agents. Not deciding faster. Making decisions take effect.

From sentence to plan

“We need a Company Brain.” The sentence from the start does not describe a software gap. It describes a company in which nobody has yet decided what holds and who maintains it. The openness is there, 3.94 out of 5. The workflows do not follow, 2.47. A Company Brain does not close this gap because it is cleverer than your people. It closes it because it takes your decisions to where people and AI agents work, with an accountable owner behind them.

The companies that get this right in 2026 will not be the ones that picked the best platform. They will be the ones that decided early who is accountable for what the company holds to be true.

Bring one important decision. In 30 minutes with Marc or Simon we show which tickets, documents and commitments contradict it today. Get to know the Company Brain

Frequently asked questions

Recommendations for the first 30 days

• Week 1: owner, decision, three topics. Name an accountable owner for the Company Brain. Pick one important decision that gets lost in execution today, and three topics that must never be out of date, such as strategy, pricing and customer commitments, each with a subject owner.
• Weeks 2 and 3: connect conversations and systems. Record leadership and planning meetings by default from now on. Connect the two or three systems where the decision is executed, such as Jira, Confluence and the code, and check what contradicts it today.
• Week 4: check, approve, write back. Write 20 real questions per topic with checked answers, approve the first corrections and write them back to the source systems. Time to value matters, and once the Company Brain is built in, things move fast: every further decision uses the same connections.

Scope & caveats

• The AI Monitor 2026 measures readiness, not ROI. The DACH analysis covers 64 companies (46 CH, 18 DE) from a master's thesis at the University of St.Gallen with Teklens as practice partner, the European pre-study 165 participants; all self-assessments, exploratory, not representative. Leaders, Followers and Laggards are k-means profiles built from readiness scores, not success groups; for 10 of the 12 Leaders the assessment came from top management. Mapping access, orientation and ways of working to the dimensions is my interpretation, not a tested intervention.
• The Menlo figure is based on a survey of around 500 decision-makers at US enterprises. The CHOIR study observed 21 people in four research groups over one month; its finding is a signal, not a measurement for companies. The details on personal AI agents reflect the vendors' announcements in autumn 2026 and can change quickly. I describe Teklens as it is today.
• The rules on ownership, labelling, knowledge hubs and the bridge are my practice rules, not study results. To go deeper: Nufar Gaspar describes many of them in her operator guides to the Company Brain and the personal brain (see sources).

Sources

Every external figure and quote in this piece – linked so you can verify it.

1. 1.Probst (2026), AI Monitor 2026, DACH-Analyse: 64 Unternehmen, 19 Faktoren, sechs Dimensionen, Skala 1–5 ↗ – Master's thesis at the University of St.Gallen in collaboration with Teklens; k-means profiles Leaders, Followers, Laggards.
2. 2.AI Monitor 2026, europäische Vorstudie (165 Teilnehmende) ↗ – Tool use: 163 valid responses.
3. 3.Menlo Ventures (2025), «2025: The State of Generative AI in the Enterprise» ↗
4. 4.Lee et al. (2026), «CHOIR: A Chatbot-mediated Organizational Memory Leveraging Communication in University Research Labs», CHI 2026, arXiv:2509.20512 ↗
5. 5.Meta Newsroom (8. September 2026), «Introducing Muse» ↗
6. 6.OpenAI (29. September 2026), «Introducing dots» ↗
7. 7.Apple Newsroom (September 2026), «Siri AI, a profoundly more capable and personal assistant, is here» ↗
8. 8.xAI, «Introducing Grok Bot» ↗
9. 9.TechCrunch (24. August 2026), «Instinct's powerful AI assistant is raising privacy and security concerns» ↗
10. 10.Bloomberg (28. September 2026), «Manus Expands AI Tools in Renewed Push Into Agent Market» ↗
11. 11.Lindy ↗
12. 12.OpenClaw (GitHub, Peter Steinberger) ↗
13. 13.Nous Research, Hermes Agent (GitHub, MIT-Lizenz) ↗
14. 14.Nufar Gaspar (2026), «The Company Brain (and the Bridge): An Operator's Guide for Leaders» ↗ – Further reading.
15. 15.Nufar Gaspar (2026), «Your Personal Brain: A Practical Guide for Leaders, Practitioners, and Enthusiasts» ↗ – Further reading.

Terms around the Company Brain

Every term has its own glossary page, with a definition, a practical example and the way back to this guide.

• Company Brain
• Enterprise Context Management
• Meeting Companion
• Playbook
• Organisational memory
• Single Source of Truth
• Strategy execution gap
• AI Readiness
• Second Brain
• AI memory
• AI agent
• MCP
• Human in the Loop
• AI governance
• Shadow AI
• LLM data controls
• Knowledge management with AI
• Context Engineering

Keep reading in the AI Leadership Lab

Related deep dives – from the same pillar and the adjacent phases.

Matching Activities from the Playbook

From the article straight into practice: these Activities put the concepts to work with Teklens.