Skills are not the same as capability

Much of the current response to AI has focused on skills.

There is an understandable logic to this. As tools change, people are expected to learn how to use them. New courses emerge, new competencies are defined, and new forms of training are introduced to keep pace.

But there is a risk in assuming that skills, on their own, are enough.

Skills describe what someone can do in isolation. They can often be taught, assessed, and certified in relatively controlled settings. In many cases, they are also transferable — a tool learned in one context can be applied in another.

Capability is different.

Capability is expressed in context. It involves not only knowing how to do something, but when to do it, why it matters, and how it fits within a broader process or decision. It includes judgment, prioritisation, and the ability to navigate ambiguity.

In environments where AI is increasingly present, this distinction becomes more important.

A person may learn how to use an AI tool effectively — how to prompt it, refine outputs, or integrate it into a workflow. But this does not necessarily mean they can use it well in situations where the stakes are unclear, the context is complex, or the consequences are significant.

The gap between skill and capability often becomes visible at precisely these moments.

For example, an AI-generated output may appear coherent and complete. A skilled user may know how to produce it efficiently. But capability is reflected in how that output is interpreted — whether assumptions are questioned, whether context is properly applied, and whether the result is appropriate for the situation at hand.

This is not something that can be fully captured through checklists or standardised assessment.

It develops through experience, but also through exposure to real situations where decisions carry weight. It requires opportunities to test reasoning, to reflect on outcomes, and to understand not only what worked, but why.

In this sense, the challenge is not simply to equip people with new tools, but to support the development of capability in environments where those tools are already influencing how work is done.

This gap also has implications beyond individuals. If organisations focus primarily on skills, they may succeed in increasing tool adoption without necessarily improving decision quality. Processes may become more efficient, but not more robust. Outputs may increase, but without a corresponding increase in confidence in how those outputs were produced.

Over time, this can lead to a quiet mismatch between apparent competence and actual capability.

Seen in this light, the question is not whether people have the right skills, but whether they are developing the capability to use those skills responsibly, in context, and under conditions that are not always predictable.

Part of the “Capability in an AI-shaped world” series.