Assess whether notices match the actual decisioning reasons (2e596e)
August 31, 2026
SITUATION HMDA data-quality manager in a credit union rolling out a special-purpose credit program has one working extract — model-reason-code mapping that does not match notices — after an underwriter chat that used coded language. If model-reason-code mapping that does not match notices cannot support notices match the actual, the only defensible Fair Lending output is hold.
DECISION HMDA data-quality manager in a credit union rolling out a special-purpose credit program must choose Remove access or reverse the item / Temporary compensating control / Approve a documented exception / Hold using model-reason-code mapping that does not match notices after an underwriter chat that used coded language.
HYPOTHESES TO TEST 1. An underwriter chat that used coded language is noise around an already-controlled CRA and Special-Purpose Programs process in a credit union rolling out a special-purpose credit program, given model-reason-code mapping that does not match notices. 2. An underwriter chat that used coded language is the event in model-reason-code mapping that does not match notices that forces Remove access or reverse the item for HMDA data-quality manager under Fair Lending. 3. Model-reason-code mapping that does not match notices shows a one-file miss after an underwriter chat that used coded language, not a CRA and Special-Purpose Programs program failure. 4. Model-reason-code mapping that does not match notices cannot decide notices match the actual yet after an underwriter chat that used coded language; hold is the only Fair Lending close a credit union rolling out a special-purpose credit program can defend.
ANALYSIS REQUIRED 1. Check HMDA coding and underwriting policy against notices match the actual. 2. Compare model-reason-code mapping that does not match notices to similarly situated files, second-review notes, and reason codes after an underwriter chat that used coded language. 3. Flag any disparate-impact table HMDA data-quality manager cannot explain from model-reason-code mapping that does not match notices. 4. For this Fair Lending CRA and Special-Purpose Programs file, read model-reason-code mapping that does not match notices against an underwriter chat that used coded language and write the one fact that would move notices match the actual for HMDA data-quality manager.
RECOMMENDATION Choose Remove access or reverse the item / Temporary compensating control / Approve a documented exception / Hold on this Fair Lending / CRA and Special-Purpose Programs packet (model-reason-code mapping that does not match notices after an underwriter chat that used coded language). The follow-on CRA and Special-Purpose Programs action is what HMDA data-quality manager does next: implement the option, assign an owner, and log the missing fact.
Explore more
More Fair Lending prompts
- Assess whether a special-purpose program is well designed or a pretext
- Assess whether a redlining pattern exists after controls (53c70f)
- Assess whether the exam response should concede a finding (ab398d)
- Assess whether a redlining pattern exists after controls (a8a83c)
- Assess whether a special-purpose program is well designed or a pretext
Explore related decision areas
- Assess whether loss development requires a rate or a restriction after a CATInsurance Underwriting
- Whether product recall exposure is priced or excludedInsurance Underwriting
- Assess whether the board has been accurately briefed (ff7dc2)AI Governance
See governed multi-model AI on your own prompt
Compare GPT-5, Claude, and Gemini side by side, with human review and a decision record built in.

