Risk AssessmentCritical riskComparison recommended

AI CRA Strategic Assessment Playbook

A bank with $8B in assets received a 'Needs to Improve' CRA rating in its last examination. The bank is in its 3rd year of a merger application review. The Federal Reserve has conditioned merger approval on an improved CRA rating. The next CRA exam is scheduled in 14 months.

When to use this playbook

  • Use this playbook when the decision looks like the situation above: A bank with $8B in assets received a 'Needs to Improve' CRA rating in its last examination.
  • It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "CRA Strategic Assessment".
  • Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.

What you'll need

  • Prior CRA examination report with specific findings
  • Current CRA activity register (loans, investments, services)
  • Assessment area demographics and LMI census tract map
  • Peer institution CRA performance data
  • Federal Reserve merger condition letter

Attachments: Multiple attachments (Spreadsheets, Documents)

The Prompt

You are a CRA officer developing a strategic improvement plan to achieve a Satisfactory CRA rating within 14 months. I am attaching:

Work only from the attached source files. If a conclusion is not supported, say so.

Produce:
1. Identify the specific CRA test failures from the prior examination: Lending Test, Investment Test, Service Test—and what the examiner said was missing.
2. Calculate the LMI lending volume needed to move from 'Needs to Improve' to 'Satisfactory' under the Lending Test, based on peer comparisons and examiner guidance.
3. Develop the 14-month investment and service activity plan: what CDFI investments, small business lending targets, and branch/ATM service commitments will move the rating.
4. Identify the highest-credit-value activities: what counts most under the new CRA rules and what generates the most examiner credit per dollar deployed.
5. Tell me what to present to the Federal Reserve as evidence of good-faith progress before the merger decision and what the risk is if the exam falls short.

Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.

What to expect

  • Prior exam failure analysis by test
  • LMI lending volume target calculation
  • 14-month activity plan with credit value estimates
  • High-value activity prioritization
  • Federal Reserve progress reporting strategy and merger risk assessment

Review before you act

  • Validate this output against source files before relying on it: Identify the specific CRA test failures from the prior examination: Lending Test, Investment Test, Service Test—and what the examiner said was missing.
  • Validate this output against source files before relying on it: Calculate the LMI lending volume needed to move from 'Needs to Improve' to 'Satisfactory' under the Lending Test, based on peer comparisons and examiner guidance.
  • Validate this output against source files before relying on it: Develop the 14-month investment and service activity plan: what CDFI investments, small business lending targets, and branch/ATM service commitments will move the rating.
  • Validate this output against source files before relying on it: Identify the highest-credit-value activities: what counts most under the new CRA rules and what generates the most examiner credit per dollar deployed.
  • 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 CRA Strategic Assessment, 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 prior exam failure analysis by test; lmi lending volume target calculation; 14-month activity plan with credit value estimates; high-value activity prioritization. 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.

Fair LendingCRA and Special-Purpose ProgramsRisk AssessmentCriticalMultiple attachments

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.