AI Playbook for Fair Lending Examination Preparation
A bank has received a 90-day notice of a joint OCC/CFPB fair lending examination. The bank has a mortgage portfolio, auto loan portfolio, and small business lending program. Prior examinations noted concerns about pricing discretion. The compliance team has 90 days to prepare.
When to use this playbook
- Use this playbook when the decision looks like the situation above: A bank has received a 90-day notice of a joint OCC/CFPB fair lending examination.
- It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "Fair Lending Examination Preparation".
- Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.
What you'll need
- Prior examination reports (2 most recent cycles)
- Current mortgage, auto, and small business loan data (HMDA LAR and equivalent)
- Compliance management system documentation
- Fair lending training records
- Examination notification letter and stated scope
Attachments: Multiple attachments (Spreadsheets, Documents)
The Prompt
You are a compliance officer developing a fair lending examination preparation plan for a joint OCC/CFPB examination. I am attaching: Work only from the attached source files. If a conclusion is not supported, say so. Produce: 1. Identify the top 5 areas the examiners are most likely to focus on based on prior findings and current portfolio data—map the risk to specific examiner test procedures. 2. Conduct a self-test of each portfolio: run the same analysis the examiners will run on pricing, approval rates, and geographic distribution. 3. Identify any findings that the self-test reveals that should be remediated before the examination starts—early remediation is mitigating. 4. Prepare the document request response: what records the examiners will ask for on Day 1 and how to organize the production. 5. Tell me the examination management strategy: who speaks for the bank, what we provide proactively, and what we do not provide without a specific request. Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.
What to expect
- Top 5 examination focus area predictions
- Self-test results by portfolio
- Pre-examination remediation priority list
- Day 1 document production organization
- Examination management strategy
Review before you act
- Validate this output against source files before relying on it: Identify the top 5 areas the examiners are most likely to focus on based on prior findings and current portfolio data—map the risk to specific examiner test procedures.
- Validate this output against source files before relying on it: Conduct a self-test of each portfolio: run the same analysis the examiners will run on pricing, approval rates, and geographic distribution.
- Validate this output against source files before relying on it: Identify any findings that the self-test reveals that should be remediated before the examination starts—early remediation is mitigating.
- Validate this output against source files before relying on it: Prepare the document request response: what records the examiners will ask for on Day 1 and how to organize the production.
- Confirm every cited figure, date, counterparty, or requirement against the attached originals — models compress and can drop a qualifier.
- Treat disagreement between models as a review item, especially on classification, materiality, and recommended next action.
- Do not authorize an operational, clinical, legal, credit, or enforcement action solely because the models agree.
Why compare models on this
For Fair Lending Examination Preparation, running the same attachments across independent models is useful because the hard part is classification and completeness, not fluency. The workflow is already designed to surface top 5 examination focus area predictions; self-test results by portfolio; pre-examination remediation priority list; day 1 document production organization. Those are comparison artifacts — they only exist if more than one model runs. Control specifications, geographic market definitions, and 'similarly situated' calls routinely diverge. Model disagreement is a signal to re-cut the file review, not to publish a single p-value.
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.

