Enterprise AI has hit a problem: knowledge about work exists only in people’s heads

Companies invest billions in corporate AI, but face a fundamental problem: most knowledge about how work actually functions lives not in databases and documents — but in employees' heads. And extracting this knowledge is far more difficult than it seems.

Invisible company knowledge

Formal corporate systems — CRM, ERP, Jira, Confluence — capture results and formal processes. But they almost never capture what truly matters: how an experienced manager resolved a contentious issue with a key client, why a team has circumvented a bottleneck in the pipeline in exactly this way for years, what nuances are crucial when working with a specific partner. This phenomenon is called "tacit knowledge" — a term introduced by philosopher Michael Polanyi.

Its essence: an expert often cannot explain why they make a particular decision — they simply know. An experienced cook understands by touch when dough is ready. An experienced lawyer instantly spots a problematic clause in a contract.

AI without such context remains blind regardless of how powerful its architecture is.

Why this is hard to digitize

Corporate AI runs into this problem from several angles:

• Articulation: employees often cannot describe what they do — they act by intuition developed over years
• Volatility: informal knowledge constantly updates under the influence of new events, while documentation falls behind
• Incentives: an employee has no direct motivation to explain exactly how they work — it is their competitive advantage within the company
• Context: the same rule works completely differently depending on the client, deadline, and team composition
• Scale: in a large organization, thousands of knowledge holders exist, and each has their own version of the same processes

Even a powerful language model trained on the company's official documents will give advice that works on paper but diverges from reality. This is what Bloomberg calls the main unsolved challenge of corporate AI today.

What companies are already trying

Several approaches are gaining traction. Some companies launch "knowledge mining" programs — series of structured interviews with key experts, recordings of which are then used for fine-tuning internal models or populating corporate knowledge bases. Labor-intensive, but yields high quality.

Others implement process mining tools: they automatically analyze logs from ERP and CRM systems and reconstruct actual work processes — what actually happens, not what is written in regulations. The gap between them can be striking.

Finally, a whole class of software has emerged — AI "shadows," which observe an employee's actions in the background and gradually build a model of their behavior. This raises serious questions about privacy and ownership rights of knowledge accumulated by a person over years of work.

"The real challenge is not training a model on company data.

The challenge is first extracting that data from people."

What this means

Corporate AI is constrained not so much by technology as by organizational anthropology. Companies that learn to systematically extract and digitize their employees' informal knowledge will gain a real competitive advantage — not from more intelligent models, but from better data about how their business actually works. This shifts competition from the question "whose AI is more powerful" to "who knows themselves better."