The Knowledge AI Can’t See

Tacit Knowledge and the Illusion of Understanding

There’s a lot of noise right now about AI in the workplace.

Some of it is excitement. Some of it is fear. Most of it sits somewhere in between, framed around productivity, efficiency, scale, and speed. The underlying assumption is often the same: that work is made up of knowledge that can be captured, codified, automated, and transferred.

I’m broadly optimistic about AI. I see its potential every day when it’s used well. But my interest isn’t in debating whether AI is good or bad. It’s in understanding what happens to systems, and to people, when powerful tools are introduced into environments that already struggle with memory, alignment, and care.

That’s where tacit knowledge becomes impossible to ignore.

Tacit knowledge is not the kind of knowledge that sits neatly in documents, frameworks, or dashboards. It’s not easily articulated or transferred. It lives in judgment, context, timing, and relationship. It shows up in knowing when to push and when to pause. In understanding how a decision will land, not just whether it is technically correct. In recognising early signals before they become metrics.

Most organisations run far more on tacit knowledge than they realise.

It’s held by people who know how things really work. Who understand the informal dependencies, the workarounds, the risks that don’t show up on paper. Who carry institutional memory not as records, but as lived experience. Often, these are the same people who care deeply about outcomes and feel the weight of responsibility when systems don’t quite hold.

This matters because AI changes what organisations pay attention to.

As work becomes more automated, more legible, and more measurable, there’s a subtle shift in what is valued. What can be seen, tracked, and optimised begins to stand in for understanding itself. Efficiency improves. Output increases. Expectations rise. Demand expands to fill the new capacity.

At the same time, reliance on tacit knowledge doesn’t disappear. It increases.

The paradox is that as systems become more efficient, they lean even more heavily on what cannot be automated. Judgment. Sense-making. Care. Context. The human ability to hold complexity and ambiguity. But because these things are harder to see, they are often treated as background conditions rather than critical infrastructure.

AI doesn’t create this dynamic. It accelerates it.

In organisations already prone to short memory and repeated crisis, this acceleration can be dangerous. When people leave, what is lost is not just headcount or capability. What leaves quietly is continuity. Nuance. The knowledge of how past failures were navigated, why certain decisions were made, and where the real pressure points sit.

This loss is rarely immediate or dramatic. Systems continue to function. On the surface, things appear stable. But over time, brittleness sets in. Decisions feel less grounded. Mistakes repeat. The same interventions are required again and again, without a clear sense of why.

This is where the themes of the last two pieces converge.

People who hold tacit knowledge are often the same people caught in the triangulation of care, system, and purpose. They care about the work and the communities it serves. They operate inside systems that struggle to change. They carry memory, context, and responsibility that the organisation does not formally recognise.

When those people are stretched, injured, or leave, the system doesn’t just lose labour. It loses its ability to learn.

AI can either amplify this loss or help surface it.

Used well, AI can free people from low-value work, create space for deeper thinking, and support better decisions. But only if organisations remain clear-eyed about what AI cannot see, cannot hold, and cannot replace. Tacit knowledge doesn’t become less important because technology improves. It becomes more so.

The real risk isn’t that AI removes humanity from work. It’s that it convinces organisations they understand their systems better than they actually do.

If we mistake legibility for understanding, and efficiency for wisdom, we risk building systems that optimise relentlessly while forgetting what makes them resilient in the first place. Not grit. Not coping. But memory, judgment, and care held by people who know how things really work.

The question, then, isn’t whether AI belongs in our organisations. It already does.

The question is whether we are designing systems that recognise, protect, and learn from tacit knowledge, or whether we are accelerating its erosion while believing we’ve finally solved complexity.

If you’re exploring AI in your organisation and want to think more carefully about what gets amplified, lost, or misunderstood along the way, I work with leaders on system design and decision-making in moments of change then reach out via info@dialecticalconsulting.com.au or contact me via LinkedIn.