What does it mean for an agent to decide whether the supplied documents are enough?

Before an agent can approve, route or book anything, it has to answer a prior question: do the supplied documents support the next step? In a loan application, an insurance claim, a utility connection or a supplier onboarding, a case handler checks the bundle against a requirement list, notices what is absent, stale or inconsistent, and either proceeds or asks for specific items.

This is different from extraction. The handler is not reading fields; they are judging sufficiency. A payslip may be present but older than the policy allows. A proof of address may be present but name a different entity. A signed form may be missing a page.

For agent developers the workflow matters because errors are asymmetric. Proceeding on thin evidence creates downstream rework, exceptions and sometimes unsound decisions. Asking for too much irritates customers and stalls queues. The agent also has to write a request specific enough for a person to act on, and then handle a reply that arrives only partly complete.

Training and evaluation need cases where the right answer is 'not yet', where the requirement depends on case type, and where the gap is only visible by comparing two documents. MirrorFirm generates synthetic material shaped by direct observation of how handlers make these calls. Buyers should still inspect grounding and disclosure risk in any sample.