Capability map

Capability Map

A structured view of what an AI system can actually do — what skills it has, what tools it can call, what data it can read, what decisions it can make — so the team and the users share the same model of its scope.

Operating principle

Production AI is not a prompt. It is a system of context, tools, permissions, traces, evals, and feedback loops.

What it is

A capability map is a living document (often code-backed) that describes everything an AI system is capable of doing: which intents it recognizes, which tools it can invoke, which knowledge sources it reads from, which actions it can take, and what guardrails apply to each. Think of it as the org chart for the AI's responsibilities.

Why it matters

Without a capability map, the team building the AI and the users consuming it disagree on what it should do. Users ask for things outside its scope; the team adds features users don't notice; expectations and reality drift apart. A capability map keeps both sides honest about scope.

How it works

Each capability is a typed entry with: name, description, the workflow or tool that implements it, the data sources it reads, the actions it can take, the approval gates it triggers, and the eval cases that prove it works. The map is queryable — agents can use it to introspect, users can use it to set expectations, and ops can use it to deprecate cleanly.

Related resources