Capability is not authority
A design principle for AI agents: being technically able to perform an action does not mean the agent should be permitted to perform it. Delegation needs gradients of authority.
An AI agent that can read a calendar, move a meeting, draft an email, send it, edit a customer proposal, place an order or update a system of record has many capabilities. That does not mean all of those actions should carry the same authority. Capability describes what the agent can do. Authority describes what it is allowed to do, on whose behalf, under what conditions.
Delegation therefore needs gradients. Some actions can run freely, some need a recorded reason, some need confirmation, and some should always go back to a person. Delegation turns permissions into product design.
Why it matters
As agents become persistent and always-on, the gap between "can" and "may" is where most real risk lives. It is the same distinction that appears in decision systems: confidence is not permission. A model being 90% sure does not decide what it may do about it; a threshold policy does.
Pair this principle with guardrails, clear identities and permissions, and honest assessment of capability readiness.
Read more in OpenAI Dots is a test of whether AI can carry a goal, not just complete a task.
Related terms
Persistent delegation
Handing an AI agent a goal it keeps carrying over time — remembering context, noticing relevant events and acting across applications — rather than completing a single prompted task.
Capability readiness
How ready an organization's actions are to be used by AI agents: whether each action is a real business capability, exposed cleanly, with clear permissions, side effects and recovery.
Threshold policy
The explicit rules that turn a model's probability or score into an action — reject, send to a human, or act autonomously — owned and reviewed separately from the model itself.
Guardrails
Controls placed around an AI system — input and output checks, permission limits, review steps and escalation rules — that keep its behaviour within acceptable bounds.