Marketing teams today have access to the best AI tools in history, and yet concepts take a long time to develop, agreements get lost in inboxes, and new team members spend months asking about things the company already knows. I see this regularly, and to put it bluntly: the problem rarely lies in the tools. It lies in where the organization’s knowledge lives and how little of it is visible from the perspective of a single department.
What Is an Organizational Second Brain, and Why Isn’t It About Notes?
An organizational second brain is a shared, curated repository of a company’s knowledge and skills, used by both people and AI agents. It stores what the company knows: strategy, agreements, decision history, market knowledge. It also stores what the company can do: documented, repeatable procedures for getting work done. And it defines who decides what counts as true within it.
The concept itself has a longer history than the AI trend suggests. Niklas Luhmann, a sociologist and one of the most productive scholars of the 20th century, treated his paper slip box as a conversation partner and credited it with much of his body of work1. In his book “Building a Second Brain”, Tiago Forte brought this idea into the digital world and defined a second brain as an external, centralized repository of what you learn2. In 2026, Andrej Karpathy added a third link by describing the “LLM wiki” pattern: a large language model (LLM) not only answers questions but also writes and maintains knowledge stored in files, like a compiler that turns raw sources into an organized, interlinked encyclopedia3.
This brings us to the heart of the matter. T. Forte described a personal system, the private memory of one individual. An organizational second brain is something different: it takes the same idea to the company level, where knowledge has to be shared, up to date, and trustworthy for many people at once. The difference is best described through three layers that a traditional corporate wiki never had all together:
| Layer | What it covers | What a traditional wiki lacks |
|---|---|---|
| Knowledge | strategy, agreements, market, competition, decision history | the knowledge is there, but you have to search for it manually |
| Skills | documented procedures: how we write a brief, how we report on a campaign, how the brand sounds | missing; a wiki describes the work but doesn’t do it |
| Governance | who is the source of truth, who approves changes | missing; everyone edits everything, so no one trusts anything |
That’s why it isn’t about notes. It’s about a system in which knowledge does the work.
What Does Invisible Knowledge Cost a Company?
This problem is older than AI. As early as 2012, the McKinsey Global Institute estimated that the average knowledge worker spends nearly 20% of the workweek searching for internal information or tracking down colleagues who can help with a task4. A 2018 study by Panopto and YouGov adds a second dimension: 42% of the knowledge needed for a given role exists only in one person’s head, and employees lose an average of 5.3 hours a week waiting for information from colleagues or recreating knowledge that already existed in the company5. When that one person leaves, the knowledge walks out the door with them.
AI was supposed to fix this, but it won’t do so on its own. According to an EY study from April 2026, 51% of medium-sized and large companies in Poland say AI brings them real benefits, yet only 9% rate their data infrastructure as fully ready; 77% of organizations plan to increase their AI spending6. The global picture is similar: 88% of companies regularly use AI in at least one business function, and yet in no single function do more than 10% of respondents report scaling AI agents in earnest7. The tools are everywhere; the value still gets stuck somewhere. In my view, it gets stuck exactly where the language model has no access to the company’s organized knowledge.
One caveat before we go further. A second brain does not replace people. It takes over searching, compiling, and first drafts; judgment and decisions stay with the team.
Organizational Memory: Knowledge That Finds Itself
The first level of maturity is memory. Knowledge stops living in people’s heads, inboxes, presentations, and the archives of former employees; it moves into a single repository where it surfaces on its own, in the context of the task at hand. You’re preparing a campaign concept for a client in the furniture industry, and the system brings up the communication strategy, last quarter’s agreements, and the lessons from a campaign that didn’t work two years ago. Not because someone remembered, but because no one had to.
In practice, it works well to divide resources into three types:
- the organization’s internal knowledge,
- external knowledge about the market and competitors,
- skills, which we’ll come back to in a moment.
For the marketing department, this memory has one more, rarely noticed consequence. An organizational second brain stores knowledge about the entire go-to-market: business goals, sales, the product, pricing, and what customer service is hearing. Marketing stops looking at the market from inside its own silo; a campaign is built with awareness of the quarter’s sales targets and the product’s positioning, not just the brief.
Who Is the Source of Truth (Governance)
Without this layer, a repository quickly loses its structure and clarity, even if its interface looks orderly. In an organizational second brain, you can designate people who are the source of truth in their domains, as well as people who approve knowledge prepared by others. There are contributors and there are curators; established knowledge is kept separate from proposals. It’s less glamorous than automation, but governance is exactly what distinguishes organizational memory from a folder full of files.
How Does Knowledge Turn into Skills?
The second level is the point at which the second brain stops being just a source of knowledge and becomes a source of skills. A skill is a documented, repeatable procedure for doing a piece of work to the company’s standard: how we write a brief, how we build a campaign report, how our language sounds, what we never do in our communication. The difference is easy to remember: a wiki tells you how we write briefs; a second brain writes the brief the way we write it.
This level also resolves the apparent paradox of speed versus quality. Intuition says that faster means sloppier. Yet skill-based automation runs a process according to the best known version of the procedure, the same way every time. People have bad days, rush, forget one step out of twelve; a skill doesn’t. Quality goes up not because AI is smarter than the team, but because the company stops paying a tax on inattention.
Day-to-day content work rests on this layer. Documented brand communication guidelines, the brand’s strategy, and a record of what has worked mean that each new concept is built on a foundation rather than from scratch. You can build entire pieces of content on it, as well as video scripts, campaign concepts, and ad creatives that meet platform specifications.
The Second Brain in Company Projects and Processes
The third level of maturity goes beyond answering questions: the second brain starts working within processes.
The first mechanism is project work inside the concept itself. The team creates ideas, develops them, and builds on one another’s work, and the course of the project, meaning what worked and what we want to avoid, feeds back into the organization’s knowledge. This creates a flywheel: the project draws on knowledge, the project builds knowledge, and the next project starts from a higher level. In the traditional model, project knowledge dies in the archive; here, it becomes input.
The second mechanism is integrations. Connecting to email, kanban boards, video conferencing tools, and the CRM (customer relationship management) system works in two directions. Aggregation: the system sees an email, a meeting note, a Trello card, and a sales opportunity as a single context, so it can answer a question about a project’s status without opening four different tools. Transfer: it drafts a reply to an email based on the history, creates a task in Jira following agreements made in a meeting, and checks whether materials from a subcontractor match the order that was sent. That last example sounds mundane, and that’s exactly why I like it; it’s precisely the kind of work no one on the team wants to do by hand.
The direction is, in fact, broader than our own practice. In its 2025 Work Trend Index report, Microsoft describes the birth of the “Frontier Firm”, an organization built around human-agent teams; 82% of leaders surveyed said they planned to deploy AI agents within 12–18 months8. In this setup, an organizational second brain is the knowledge infrastructure without which such a team has nothing to work with.
Is an Organizational Second Brain a Major IT Implementation?
No. And this is perhaps the most underrated feature of the whole concept.
In the form in which we implement it, a second brain is a structure of markdown files, that is, plain text files that a person can read and any language model can understand. A. Karpathy summed it up in a phrase that’s hard to beat: “Obsidian is the IDE; the LLM is the programmer; the wiki is the codebase”. This structure works with any platform that supports projects and agents, and, in the most extreme case, with a model hosted on the company’s own infrastructure when security requirements rule out the cloud.
The choice of model and platform follows from four criteria: compliance guidelines, the functionality the model needs to deliver, how advanced the users are, and the budget for implementation and tokens. The configuration follows from the requirements, not the other way around; it’s not rocket science, but it does require well-organized data.
If the cost isn’t the technology, then what is it? Persistence and time: gathering scattered data, structuring it, and training the team to use the tool every day. From the end user’s perspective, it’s often simply a slightly more advanced way of working with a language model; from the organization’s perspective, it’s a change of habits, and changing habits always costs more than licenses.
Two Layers of Work: Strategic and Operational
At Yetiz, we work with this concept on two layers, and I recommend this split to anyone wondering where to start.
The strategic and business layer builds knowledge about the client: their needs, goals, and communication strategy, and it helps define tactics. This is where content concepts grounded in sales goals and product positioning are created, along with complete content in the brand’s voice, video scripts, and campaign concepts. Importantly, this kind of instance can run in the client’s environment, with access to their internal resources; it then builds knowledge, reports, and analyzes based on context that no external agency can see.
The tactical and operational layer supports building and optimizing campaigns, primarily by synchronizing and analyzing data from multiple sources and combining it into a single, coherent picture of the campaign as a whole. This is also where the work I described above lives: email replies, tasks, and checking that materials match the order.
I won’t give you a table of efficiency gains in percentages here. What you can see with the naked eye is the quality of the work. Many processes are structured, and the brains make sure they are carried out and analyze data of a complexity that can no longer be sensibly handled in a spreadsheet; in practice, they act like a shared manager who knows every agreement and never has a bad day. They oversee processes, not people. Meanwhile, the team does what it is really needed for: making data-driven decisions and drawing conclusions, instead of spending energy on operational tasks and risking mistakes by doing them manually.
How to Implement an Organizational Second Brain
The order matters. I see five steps, and the last one determines everything.
- Define the purpose. It will be different in an agency, in a marketing and sales department, and in e-commerce. Without a purpose, the repository turns into a storage closet.
- Define the working model. Who uses it, who is allowed to contribute, who is a passive user; what rules govern how projects are run in it. This is the moment when governance is born.
- Determine the starting knowledge and set of skills. The system has to show its value from day one, so at launch it needs a minimum of knowledge and a few skills that address the team’s real, weekly needs.
- Start with a pilot. One team, one process, a few weeks. Rolling it out everywhere at once is the fastest way to have the tool abandoned, second only to having no owner.
- Last but not least: appoint an owner. A person who oversees the implementation and then manages its development. The value of this platform lies in its continuous adaptation to needs; skipping this role will cause the concept to die within a few weeks, because it will stop keeping pace with the organization.
To be fair, this implementation also has costs that don’t show up in the budget. The initial knowledge inventory can be tedious, curation discipline takes time from the very people who have the least of it, and some team members are initially reluctant to share knowledge, because knowledge held in one head has so far served as job insurance. It’s worth naming this resistance at the outset instead of pretending it won’t happen.
Second Brains That Talk to Each Other
Finally, the direction I find the most interesting. If the strategic layer can run in the client’s environment and the operational layer at the agency, the natural next step is communication between them.
Flowing down, from the strategic and business layer to the operational layer, are business data or the signals derived from it, along with strategic assumptions. Flowing back up are insights and reports measuring the effectiveness of activities carried out under those strategies, as well as recommendations on strategy and business assumptions, including product ones, that come directly from the campaigns. The campaign stops being the end of the chain; it becomes a source of strategic knowledge. And sensitive data never leaves the client’s ecosystem in the process, because what travels between the brains are signals and insights, not raw data. The classic conflict of “the agency wants full data, the client doesn’t want to share it” simply disappears.
Mechanically, this sounds less impressive than it looks in action. Today, this communication takes place through a shared file space with a defined structure, to which each brain publishes the final versions of documents; it’s quick to implement and doesn’t require any exotic infrastructure. In parallel, we’re working on an approach in which instances talk to each other via MCP (Model Context Protocol), an open standard for connecting AI systems to data and tools.
Imagine a morning when the team running the campaigns receives specific guidance from the client’s second brain based on the previous week’s sales data, and the client’s strategy is updated with insights from the campaigns before anyone has written a report. This isn’t a distant vision; it’s the direction in which this architecture is naturally heading.
Key Takeaways
- An organizational second brain has three layers, not one: the company’s knowledge, documented skills, and governance, meaning who is the source of truth. A traditional wiki has only the first layer, and that’s why it dies.
- The problem this concept solves is older than AI: as early as 2012, McKinsey estimated that searching for information takes up nearly 20% of the workweek, and according to Panopto, 42% of role-specific knowledge exists in only one person’s head.
- AI tools alone are not enough, because only 9% of Polish companies rate their data infrastructure as ready; the value of AI gets stuck where the model has no access to the company’s organized knowledge.
- The biggest change happens at the transition from knowledge to skills: a wiki tells you how we write briefs, while a second brain writes the brief the way the company writes it.
- This is not a major IT implementation: at its core is a structure of plain text files, and the real cost is the persistence and time needed to gather knowledge, structure it, and train the team.
- Success depends on the owner, the person responsible for developing the system after implementation; without them, the concept dies within a few weeks because it stops keeping pace with the organization’s needs.
- The direction of development is communication between brains: signals and strategic assumptions flow down, insights and recommendations flow back up, and sensitive data never leaves the client’s ecosystem.
FAQ
How Is an Organizational Second Brain Different from a Corporate Wiki or Intranet?
A wiki stores descriptions and waits for someone to look for them. An organizational second brain finds knowledge on its own in the context of the task, performs work according to documented procedures (skills), and has a governance layer that determines what the current truth is. In short: a wiki documents the work; a second brain helps do it.
Does a Second Brain Require Sending Company Data to an External AI Provider?
Not necessarily. At its core is a file structure over which the company has full control, and the choice of language model remains open, from cloud services to models hosted on the company’s own infrastructure. The configuration is determined by compliance guidelines, the required functionality, how advanced the users are, and the budget.
Where Should a Marketing Department Start Building a Second Brain?
With a purpose and a pilot, not with technology. Choose one team and one repeatable process, such as creating content concepts or reporting on campaigns, gather the starting knowledge, define the first two or three skills, and appoint an owner responsible for its development. You can build out the rest once the tool has shown its value.
- N. Luhmann, “Kommunikation mit Zettelkästen. Ein Erfahrungsbericht” (1981); English translation: “Communicating with Slip Boxes”, https://luhmann.surge.sh/communicating-with-slip-boxes (accessed August 6, 2026). ↩︎
- T. Forte, “Building a Second Brain”, Atria Books, 2022. ↩︎
- A. Karpathy, “llm-wiki.md”, April 2026, https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f (accessed August 6, 2026). ↩︎
- McKinsey Global Institute, “The social economy: Unlocking value and productivity through social technologies”, July 2012, https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-social-economy (accessed August 6, 2026). ↩︎
- Panopto, “Workplace Knowledge and Productivity Report”, 2018; study conducted with YouGov among 1,001 employees of U.S. companies with more than 200 employees, https://www.prnewswire.com/news-releases/inefficient-knowledge-sharing-costs-large-businesses-47-million-per-year-300681971.html (accessed August 6, 2026). ↩︎
- EY Poland, “Jak polskie firmy wdrażają AI?” [How are Polish companies implementing AI?], 3rd edition, April 2026; study of 497 medium-sized and large companies in Poland, https://www.ey.com/pl_pl/insights/ai/raport-ey-jak-polskie-firmy-wdrazaja-ai-gc-fy26 (accessed August 6, 2026). ↩︎
- McKinsey & Company, “The State of AI in 2025: Agents, innovation, and transformation”, November 2025, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai (accessed August 6, 2026). ↩︎
- Microsoft, “2025 Work Trend Index Annual Report: The Year the Frontier Firm Is Born”, April 2025; survey of 31,000 workers in 31 countries, https://news.microsoft.com/source/emea/features/microsofts-2025-work-trend-index-report-reveals-the-rise-of-the-frontier-firm-marking-a-new-era-of-workforce-dynamics/ (accessed August 6, 2026). ↩︎
