What is an escalation to a senior reviewer, and why does it matter for evaluation?

In a credit team, every reviewer holds a delegated authority: a limit on exposure, product type or risk grade beyond which they cannot approve alone. When a case exceeds that limit, the reviewer must stop, package the file and refer it to someone with higher authority.

The work is more than a status change. The reviewer writes a rationale, attaches the evidence relied on, names the policy clause that triggered the referral, and chooses the right next approver. The senior reviewer may approve, decline, approve with conditions or send it back for more information. Each outcome changes the case state and the audit trail.

This matters for evaluation because agents fail at it in quiet ways. They approve just inside their own limit by splitting an exposure, route to the wrong tier, drop context in the handover, or treat a return-for-information as a final decline.

Evaluation teams need cases where the correct answer depends on reading a policy table, an exposure aggregate and a case history together. A generated dataset can hold those relationships explicitly, so a score reflects whether the agent respected the boundary, not whether it guessed a plausible decision.

The material is generated, and its grounding in how reviewers actually work should be inspected in a sample before use.