Which records and relationships can a sufficiency dataset or task pack contain?

A release for this workflow can contain more than documents. The useful unit is a case: a set of linked records that lets an agent make, and be scored on, a sufficiency decision.

A dataset could include:

- **Case records** with type, channel, opening date, status history and the parties involved. - **Requirement sets** by case type and policy version, including acceptable substitutes, age limits and conditions. - **Document records** with type, issuer, dates, completeness flags and generated content. - **Decision records** stating whether the file is sufficient, the gaps found and the reasoning. - **Request records** with the message sent, the items asked for and the reply that followed. - **Handover notes** passed between intake, assessment and exceptions staff.

A task pack adds prompts, expected outcomes and scoring rules. A simulated environment adds mutable state: documents arrive over time, requests receive partial replies, and policy versions change mid-case.

Relationships matter most. A gap links to a requirement, a requirement to a policy version, a document to the case it was filed against. Conflicts such as mismatched names or dates are encoded as links between two documents rather than as a flag on one.

All of this is generated, not extracted from real records. Ask what grounding each element has, and inspect samples for anything that looks too specific to a real source.