A Company Brain often fails not because of technology, but because it starts too big. Beginning with 1,000 pages usually creates an archive that people do not use consistently. A better first step is to capture the recurring questions that cost time every month.

Why would I never start a Company Brain with 1,000 pages again?

The biggest mistake in building a Company Brain is the desire to be complete from day one. The thinking sounds reasonable: If we are finally building a digital memory for the company, we should do it properly. So old process documents are collected, wiki pages are exported, folders are scanned, PDFs are sorted, CRM notes are copied, proposal templates are uploaded and departments are asked to contribute everything they have.

After a few weeks, there is a lot of material.

But there is still no useful Company Brain.

That is the uncomfortable part. A thousand pages of company knowledge look impressive. They feel like progress. They work well in a project plan. But they rarely solve the real operational problem: employees still cannot find the one answer they need at the moment of work.

A Company Brain should not begin as a documentation project. It should begin as a relief project. The decisive question is not: “What could we document?” The better question is: “Which 20 questions cost us time every single month?”

Why is completeness so tempting in the beginning?

Completeness feels professional. It feels thorough, strategic and responsible. Many companies, especially small and mid-sized businesses, have years of accumulated knowledge: process descriptions, emails, templates, spreadsheets, customer notes, complaints, service reports, manuals and technical instructions. When a Company Brain is planned, it feels natural to capture everything at once.

But this creates complexity before value.

A large knowledge collection does not automatically become a useful knowledge system. Many pieces of content are outdated, duplicated, contradictory or irrelevant to daily decisions. Some pages were written for a one-time exception. Others describe a process that has changed. Some documents are formally correct but do not help an employee answer a concrete operational question.

Gartner describes a knowledge base as either the foundation of a successful knowledge management program or a single point of failure if structure and supporting processes are not established from the start. Source: https://www.gartner.com/en/documents/5618791

Why are 20 real questions better than 1,000 pages?

Twenty real questions force prioritization. They move the discussion from theory into daily work.

These are not questions invented in a workshop. They come from operations. Questions like: “Which pricing logic applies to this existing customer?” “Who is allowed to approve this complaint?” “Which supplier failed us last time?” “How was this special case handled before?” “Which documents are needed for this proposal?” “What did we promise the customer on the phone?” “Which privacy rule applies to this process?”

Questions like these are valuable because they reveal repetition. If the same question appears every week, it is not just a communication issue. It is a knowledge issue.

Atlassian reports that 56 percent of workers often find that the only way to get the information they need is to ask someone or schedule a meeting. Source: https://www.atlassian.com/blog/strategy/information-management

A good Company Brain starts exactly there: with questions that already burn time today.

What does the difference look like in practice?

Approach	Starting with 1,000 pages	Starting with 20 recurring questions
Goal	capture as much knowledge as possible	solve concrete search and clarification problems
Impact	much material, little daily use	faster value in real work
Risk	outdated content, overload, low adoption	limited scope, but strong focus
Maintenance	hard to control	easier to assign and review
Success metric	number of pages or migrated documents	fewer questions and faster answers
Typical mistake	building an archive	choosing a problem too narrowly
Better use	later structure and reference work	first productive Company Brain phase

The smaller approach looks less impressive. But it works better. Employees see earlier whether the system helps. The business sees earlier whether the investment creates practical value. And the Company Brain earns trust before it expands.

Why do large knowledge management projects fail in daily operations?

Large knowledge management projects rarely fail dramatically. They fade.

In the beginning, there are workshops, tools, content inventories, migration lists and ambitious goals. Then normal business returns. Proposals have to be sent. Customers wait. Projects continue. Complaints need answers. New employees need guidance. The knowledge project has to compete with everything that is urgent today.

Deloitte found that knowledge management ranks among the top three issues influencing company success, but only 9 percent of surveyed organizations felt ready to address it. Source: https://www.deloitte.com/us/en/insights/topics/talent/organizational-knowledge-management.html

That gap is visible in daily work. Companies know that knowledge matters. But they often build systems that require extra maintenance instead of immediately reducing work.

Why is search time a better business case than document volume?

When a Company Brain is discussed internally, people often talk about content volume. How many documents do we have? How many pages should be migrated? Which departments should provide material?

That is understandable, but it is not the strongest business case.

The stronger business case is search time.

Slite’s Enterprise Search Survey reports an average of 3.2 hours per week lost to searching for information. Source: https://slite.com/learn/enterprise-search-survey-findings

Even if the exact number differs by company, the pattern is familiar. Employees search emails, Teams, Slack, old proposals, folders, tickets, CRM notes and personal files. Then they ask colleagues. Then they improvise.

A Company Brain should therefore start where search time is already recurring. Not with the most elegant documentation structure. Not with the largest archive. With the questions that people cannot answer without checking several places.

How do you find the right 20 questions?

You do not find the right 20 questions only in a management meeting. You find them where the work gets stuck.

Listen in sales when proposals slow down. Listen in service when customer history is missing. Listen in accounting when invoices are delayed because agreements are unclear. Listen in project management when exceptions are discussed again and again. Listen in purchasing when supplier knowledge lives only in experience. Listen in leadership when decisions happen by memory.

A simple method works surprisingly well: collect recurring internal questions for two weeks. Not as theory, but as real cases. Who asks what? How often? Whom do they ask? How long does clarification take? Which source is missing? Which decision is delayed?

After that, the expensive questions become visible.

Which questions are usually the most expensive?

The most expensive questions are not always the most complex. They are the questions that appear often, interrupt several people or delay decisions.

Examples include:

“Which documents are required for this type of proposal?”
“Which special conditions apply to this customer?”
“What complaints has this customer had before?”
“Which suppliers are reliable for this product group?”
“Which approval is needed for this discount?”
“What was done differently in the last project?”
“Which privacy rule applies to this process?”
“Which standard response do we use for this customer type?”

If these questions return every month, they belong in the first phase of the Company Brain. Not as long articles, but as usable answers with source, context, owner and validity.

Why should you deliberately document less at the beginning?

Documenting less sounds wrong, but it is often the better decision.

Too much documentation creates maintenance work before trust exists. Employees see many entries and do not know which ones matter. Owners feel overwhelmed. Old content is migrated before anyone checks whether it is still valid. The system looks large, but uncertain.

A lean Company Brain with 20 well-maintained answers is easier to trust. Each answer can be reviewed by the right person. Each answer can be connected to a real process. Each answer can be measured: Is it used? Does it reduce questions? Does it shorten clarification? Is it kept current?

That is the difference between a knowledge archive and a working tool.

What does “start small” not mean?

Starting small does not mean thinking small. It means proving value cleanly.

A Company Brain can become broad over time. It can connect customer history, proposal logic, supplier knowledge, complaints, internal rules, process knowledge, IT documentation and compliance content. But the first step has to prove that the system helps in daily work.

Once the first 20 questions work, trust grows. Then the next 20 questions can be added. After that, knowledge domains can be grouped, data sources connected, permissions defined, AI-supported search added and workflows automated.

The difference is simple: scale from real value, not from a pile of documents.

What role should AI play in this approach?

AI can make a Company Brain much stronger, but it should not be the starting point. The starting point is the question of which knowledge the business repeatedly lacks.

Once those questions are clear, AI can help. It can search emails, documents, tickets and notes. It can find similar cases. It can prepare answer drafts. It can detect contradictions. It can show when the same topic is described differently across multiple sources.

But AI needs a reliable knowledge foundation. Gartner reported in 2026 that more than 50 percent of GenAI projects were abandoned after proof of concept due to issues such as poor data quality, inadequate risk controls, escalating costs or unclear business value. Source: https://www.gartner.com/en/articles/genai-project-failure

For a Company Brain, the conclusion is clear: do not begin with AI because it sounds advanced. Begin with real questions. Then use AI where it improves answer quality, search speed and maintenance.

How would I start today?

I would not announce a large knowledge transformation program. I would not import 1,000 pages. I would not design a perfect taxonomy before the first employee sees value.

I would begin with a simple list:

Which 20 questions cost us time every month?

Then I would treat each question individually. Who asks it? Who answers it today? Where does the information live? Are there multiple versions? Which answer is valid? Who is allowed to approve it? When must it be reviewed? Where in the workflow is it needed?

Each question becomes a small knowledge object. Not long. Not academic. Useful.

Example:

Question: “Which documents are required for this proposal type?”
Answer: structured checklist.
Source: current proposal process.
Owner: sales lead.
Valid from: date.
Review: every six months or after process changes.
Use: directly in the proposal workflow.

That is how a Company Brain starts working.

Why is this especially useful for small and mid-sized businesses?

Small and mid-sized businesses rarely have the capacity for long internal knowledge programs without immediate effect. They need relief quickly, but they also need a foundation that can grow.

That is why the 20-question approach works. It respects daily operations. It does not begin with abstract knowledge categories. It begins with real friction. It shows where people search repeatedly. It proves early whether a Company Brain actually makes work calmer.

For KrambergAI, this is central: no additional knowledge bureaucracy, but a digital company memory that supports real work. GDPR-aligned implementation, Made in Germany positioning, traceable sources and clear ownership are not decorative. They are the basis for trust.

Conclusion: What would I never do again?

I would never again start a Company Brain with volume. A thousand pages are not a success if nobody knows which answer is valid. A large import can be useful later, but not as the first step.

I would start with the most expensive recurring questions. That is where value appears fastest. That is where employees begin to trust the system. That is where knowledge management stops being an archive and becomes operational relief.

A Company Brain does not begin with the question: “What do we know?”
It begins with the question: “Which answers are missing so often that they cost us time every month?”

Sources for the statistics used

Atlassian – Stop losing time searching for the information you need
https://www.atlassian.com/blog/strategy/information-management
Deloitte – The new organizational knowledge management
https://www.deloitte.com/us/en/insights/topics/talent/organizational-knowledge-management.html
Slite – Enterprise Search Survey Report 2026
https://slite.com/learn/enterprise-search-survey-findings
Gartner – Why 50% of GenAI Projects Fail — And How to Beat the Odds
https://www.gartner.com/en/articles/genai-project-failure

FAQ

Why should a Company Brain not start with 1,000 pages?

Starting with 1,000 pages creates a lot of material, but often little practical value. Many documents are outdated, duplicated or disconnected from actual workflows. Employees do not need a large content collection first. They need reliable answers to concrete questions. That is why a smaller, measurable start is usually more effective.

Why are 20 recurring questions a better starting point?

Twenty recurring questions show where knowledge is actually missing. They come from real delays, repeated clarifications and search problems inside daily work. This connects the Company Brain directly to operational cost. Every answered question can reduce interruptions, accelerate decisions and build trust in the system.

Which questions are best for the first phase?

The best starting questions are frequent, disruptive or decision-relevant. Examples include customer-specific pricing rules, required proposal documents, complaint history, supplier experience, approval rules and privacy requirements. The key factor is not how complex a question is, but how often it costs time in daily operations.

Is starting small too slow?

No. Starting small is often faster because it creates usable results earlier. A large project needs preparation, migration, alignment and review before people see value. A focused start with 20 questions can prove practical usefulness quickly. After that, the scope can be expanded in a controlled way.

When should existing documents be imported?

Existing documents should be imported only when their purpose is clear. A mass import often creates duplicates, outdated content and confusion. It is better to use documents selectively when they answer specific questions or serve as verified sources for structured knowledge objects inside the Company Brain.

What role does AI play when building a Company Brain?

AI can search information, identify similar cases, detect contradictions and draft answers. But it should not replace the starting point. The first step is identifying which questions repeatedly cost time. After that, AI can help improve search speed, answer quality and knowledge maintenance.

How should success be measured?

Success should not be measured by the number of pages. Better metrics include fewer repeated questions, shorter search times, faster proposal preparation, fewer duplicate clarifications and higher daily usage. A Company Brain succeeds when it makes operational work calmer, faster and more reliable.

Why is this approach especially useful for small and mid-sized businesses?

Small and mid-sized businesses often lack capacity for long internal transformation projects. They need visible relief quickly. Starting with 20 recurring questions fits that reality because it focuses on real problems, avoids unnecessary bureaucracy and creates a stable foundation for a growing Company Brain.

Company Brain: What I Would Never Do Again