Why not handle this with a fixed checklist or rules engine instead of an agent?

**Why not use a fixed checklist or rules engine?**

A rules engine works when requirements are fixed and documents arrive as clean, labelled fields. Real cases are less tidy. Required items depend on case type, circumstances and policy version. Documents arrive mislabelled, combined into one file or partly legible. Two documents can each look fine and still conflict.

An agent has to combine rule-like checks with judgement. Is this letter an acceptable substitute for the form named in the policy? Is a six-week-old statement close enough, or is there a hard cutoff? Is the gap worth stopping the case for, or can the handler proceed and collect the item later?

**Does that mean rules have no place?**

They do. Many requirements are deterministic and are better checked by tools than by the model. The value for training and evaluation is at the boundary: cases where rules are silent or in tension, and where the handler's recorded decision shows which way it went.

**What does MirrorFirm add?**

Generated cases that carry the requirement set, the supplied documents, the decision and the request that followed, shaped by direct observation of how handlers work. The output is synthetic, so a buyer should still inspect how closely it is grounded in any real source and what it could disclose.