Skip to content
MirrorFirm

How can generated records represent a response that is only partly complete?

An insurance repair estimate arrives without a required labour breakdown, leaving one obligation open while the broader document request appears answered.

An insurance claims coordinator requests a repair estimate for claim `CLM-3084`. Request `REQ-6721` requires a parts schedule, a labour breakdown and a repair completion estimate. The repairer submits document `DOC-5196` before the requested response date.

The document includes parts and an expected completion date but no labour breakdown. The request's state changes to partially satisfied. When the missing component passes its deadline, policy requires the coordinator to pursue that component rather than send the entire original request again.

A generated pack could contain the request, the uploaded estimate, requirement-level completion records and the follow-up task. It could also include a draft message that mistakenly thanks the repairer for supplying everything, giving the model a concrete error to detect.

The consequence is specific: assessment of the labour charge remains blocked even though a document exists. This teaches the model that receipt and sufficiency are separate conditions. A buyer can ask for generated responses that satisfy different subsets of requirements, with explicit links showing which evidence supports each completion decision.

Represent completion at the requirement level so models learn to pursue missing content precisely.

Audience:
Model labs
Content type:
Scenario
Buyer question:
What can be generated

Content release: r-a220eb6e099b243b

More on Overdue follow up