Why do agents often get conditional approvals wrong?

**Why do agents get conditional approvals wrong?** Most agents read a status field. If it says approved, they proceed. The conditions sit in a separate list, a decision note or an attached letter, and the agent has to find them, link each to evidence, and judge whether the evidence actually meets the wording.

**Is the failure usually in reading or in judgement?** Both appear. Reading failures miss a condition buried in a decision note. Judgement failures accept a document that is attached but out of date, for the wrong entity, or below a stated threshold.

**What about conditions that change?** Conditions get varied, waived or time-limited. An agent that cached the original list can act on a condition that has since been waived, or ignore one that was added after a review.

**Why does this suit evaluation as well as training?** The correct behaviour is often to stop, ask or escalate. That is measurable. A task can score whether the downstream step was correctly held, and whether the reason the agent cites matches the condition that is actually open.

**Is the data real?** No. A MirrorFirm release is generated, informed by direct observation of how businesses handle these handovers. Buyers should inspect grounding and disclosure risk in a sample before relying on it.