Surfacing Decisions Worth a Second Look
Find the decisions that sit apart from the pattern, and hand a reviewer the specifics rather than a score.
About this use case
Quality assurance samples randomly, which means it mostly reviews ordinary files. The decisions worth a second look are the unusual ones: the approval well outside the usual profile, the decline that resembles approvals, the file decided in three minutes, the exception stack nobody escalated.
This use case compares decisions against the population and surfaces the ones that stand apart, describing what makes each unusual so a reviewer can look at the right thing.
An approval well outside the usual profile comes back with the population distribution and where the file falls; two close files decided differently are presented side by side without either being called wrong; exceptions carrying a higher authority requirement than the record shows are reported with the requirement.
Documents to consider
- Completed decisions with their attributes for the period
- Population baselines for the same attributes
- Your QA selection criteria and thresholds
- Credit policy and matrices
- Prior second-level review outcomes
The prompt
Written to be pasted straight into your AI assistant, with the documents it asks for attached. Request access below and we’ll send it over.
Files
Nothing to download here. This prompt runs against your own institution’s material — the documents it lists above.
More in Risk & Compliance
Triaging Customer Complaints
Classify an incoming complaint against regulation and internal procedure, route each issue it raises, and flag anything that needs priority escalation.
Audit Response Drafting
Draft an auditor or client response from your own control descriptions, evidence logs and prior answers, with every statement traced to an artifact.
Compliance Operations
Answer an obligation or interpretation question from approved regulatory and internal content, surfacing the exact clause that drives the answer.