User intent inference combines the current request with context such as session history, stated constraints, prior choices, and reference material. The result is a working hypothesis about what outcome the user is trying to reach.
Intent should remain revisable rather than being treated as a hidden fact. A system should track uncertainty, distinguish evidence from assumptions, and ask a targeted question when the wrong interpretation would change the result.
ELI5
User intent is what a person is really trying to achieve, which may be more specific than the words they first type. An AI system can make a careful guess from context, but it should not pretend the guess is certain.
For example, the request 'I need shoes for Rome' could mean walking shoes for sightseeing, formal shoes for an event, or something suitable for hot weather. One useful question can reveal which goal matters.
Why is user intent often unclear?
People omit details they assume are obvious, use subjective language, change goals, or begin before they know which constraints matter.
When should an AI agent ask instead of infer?
It should ask when confidence is low and different plausible interpretations would materially change the recommendation or action.
