From Knowledge Bases to Knowledge Systems

For years, we've talked about knowledge management. I think AI is about to turn that idea on its head.

The goal used to be capturing what an organization knows so people could find it later. We built wikis, documentation, research repositories, design systems, standards, playbooks, and knowledge bases. But someone still had to know the information existed, find it, and figure out when and how to apply it.

AI changes that relationship.

Knowledge no longer has to sit somewhere waiting to be discovered.

It can become an active part of how work gets done.

A research insight can inform a product decision. An accessibility standard can evaluate an interface. A design principle can shape generated code. An expert's methodology can become a reusable AI skill.

And those individual pieces of knowledge can begin to form something much larger: a knowledge system.

The shift is from storing what we know to embedding it into how the organization works.

Knowledge that actually works

Think about how much valuable knowledge exists inside any large organization.

Some of it is documented. A lot of it isn't.

It's in the designer who has spent years understanding a particular customer, the researcher who knows which questions uncover the most useful insights, the accessibility expert who immediately recognizes a problem others might miss, and the engineer who remembers why a technical decision was made five years ago.

We've traditionally thought about capturing that expertise through documentation. Write it down. Put it somewhere searchable. Teach people where to find it.

But documenting knowledge and applying knowledge are two very different things.

AI allows us to close that gap.

Instead of asking someone to find a 40-page accessibility standard and interpret how it applies, that knowledge can actively participate in evaluating an experience. Instead of hoping a designer remembers a research finding from six months ago, that insight can surface when a related product decision is being made.

The knowledge starts showing up where the work happens.

That's a much more powerful model.

From individual expertise to organizational intelligence

This also makes me think differently about expertise.

When someone on a team develops a great way of solving a recurring problem, what happens to that knowledge?

Usually they share it with a few people. Maybe they share it, document it, or incorporate it into a team process. If we're lucky, it spreads.

AI gives us another option.

We can start identifying the repeatable parts of that expertise and codifying them into reusable skills and capabilities. Those skills can be shared with other people, but more importantly, they can also be embedded into the AI systems people are using to do their work.

That creates an interesting progression.

One person's learning can become a team's capability.

A team's capability can become organizational knowledge.

And organizational knowledge can become intelligence embedded directly into AI.

Not everything should be codified. Experience, intuition, context, and human judgment will continue to matter. But there is an enormous amount of expertise inside organizations that is repeatable, teachable, and reusable.

AI gives us a new way to scale it.

This is why I think we need to distinguish between a knowledge base and a knowledge system.

A knowledge base stores.

A knowledge system participates.

It can bring the right knowledge into context, evaluate outputs against established standards, make recommendations based on previous research, and recognize when something falls outside known guidance and requires additional expertise.

Most importantly, it can connect knowledge that historically lived in separate places.

Research doesn't have to live only in a research repository. Accessibility doesn't have to live in a standards document. Design principles don't have to live on a website. Those sources can remain, but the knowledge inside them can also become part of the intelligence people and AI draw upon while work is happening.

That is a fundamentally different relationship with organizational knowledge.

A knowledge system doesn't simply retrieve what an organization knows.

It helps apply that knowledge to the decisions and outputs the organization is producing.

Knowledge needs an owner

There is also a risk here.

When knowledge moves from something people read to something AI actively applies, the consequences of getting it wrong become much greater.

If an outdated document is sitting in a wiki, it might mislead the handful of people who happen to find it. If that same outdated knowledge is embedded into an AI system making thousands of recommendations, the problem can scale very quickly.

Building knowledge systems isn't simply about connecting AI to more information. We need to know where knowledge came from, whether it's still valid, where it applies, and how it will be tested, updated, governed, and eventually retired.

And we need to be very clear about the boundary between what the system actually knows and what it is inferring.

As knowledge becomes more operational, governance becomes more important, not less.

This is a design problem too

I think UX has an important role to play here.

Designers have spent decades thinking about how people interact with complex systems: context, trust, comprehension, accessibility, feedback, and what happens when things go wrong.

Those same questions become incredibly important when designing knowledge systems.

What knowledge does someone need at this particular moment? How should AI communicate uncertainty? What evidence should accompany a recommendation? How do we make the source of the knowledge visible? When should the system make a recommendation, and when should it defer to a human expert?

Those aren't simply technical questions.

They're experience questions.

And increasingly, I think designers will help shape not only the experiences customers interact with, but also the knowledge systems that help organizations create those experiences.

That also requires clear ownership: someone has to decide what knowledge enters the system, validate where it applies, determine when it needs to change, and establish when AI can act and when a human needs to step in.

That governance model can't be an afterthought. It has to be designed into the knowledge system itself.

I also think this is where the value of our roles begins to shift.

The opportunity is to move up a level

As AI takes on more of the execution, our value becomes less about performing every individual task and more about applying judgment to the systems performing those tasks.

Without that shift, there is a real risk that we simply automate the work we do today, one task at a time, until we've slowly automated away much of the value we were providing.

Instead of only doing the work, we design and govern the systems that enable it. Instead of simply applying our expertise to a single problem, we determine how to safely codify and scale that expertise. And instead of handing judgment over to AI, we become increasingly responsible for deciding where human judgment matters most.

That requires clear ownership, governance, checks and balances, and people who remain accountable for the quality of what these systems produce. The more capable AI becomes, the more important that responsibility becomes.

The organizations that build the strongest knowledge systems won't be the ones that capture the most information or automate the most tasks. They'll be the ones that invest in the human expertise required to question that knowledge, evolve it, and know when not to use it.

That's ultimately what makes a knowledge system valuable: not how much it knows, but how thoughtfully we decide what it should know, how it should use that knowledge, and where humans remain responsible for the judgment behind it.