AI Playbook for Special Purpose Credit Program Design
A bank wants to establish a Special Purpose Credit Program (SPCP) under ECOA to expand mortgage lending to underserved communities. The bank's counsel is uncertain whether the program design meets the ECOA safe harbor requirements. The bank needs legal certainty before launch.
When to use this playbook
- Use this playbook when the decision looks like the situation above: A bank wants to establish a Special Purpose Credit Program (SPCP) under ECOA to expand mortgage lending to underserved communities.
- It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "Special Purpose Credit Program Design".
- Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.
What you'll need
- Proposed SPCP program description (eligibility criteria, benefits, geography)
- CFPB SPCP guidance (2022 advisory opinion)
- ECOA Regulation B Section 202.8 requirements
- Community needs assessment supporting the program
- HUD and FHFA SPCP guidance
Attachments: Documents (Documents)
The Prompt
You are a fair lending counsel advising a bank on Special Purpose Credit Program design under ECOA. I am attaching: Work only from the attached source files. If a conclusion is not supported, say so. Produce: 1. Assess whether the proposed eligibility criteria satisfy Regulation B Section 202.8's requirements: economically disadvantaged class, written plan, periodic reevaluation. 2. Identify any design elements that could disqualify the SPCP safe harbor: is the program genuinely targeted at the disadvantaged class or does it risk excluding the intended beneficiaries? 3. Assess the interaction with fair lending risk: could the SPCP create reverse disparate impact claims from non-qualifying applicants? 4. Review the community needs assessment: does it satisfy the requirement for a documented determination of economic disadvantage? 5. Tell me what additional documentation to create before launch and how to monitor the program to maintain SPCP safe harbor qualification. Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.
What to expect
- Regulation B 202.8 compliance assessment
- Design element risk analysis
- Reverse disparate impact risk evaluation
- Community needs assessment adequacy review
- Pre-launch documentation checklist and ongoing monitoring plan
Review before you act
- Validate this output against source files before relying on it: Assess whether the proposed eligibility criteria satisfy Regulation B Section 202.8's requirements: economically disadvantaged class, written plan, periodic reevaluation.
- Validate this output against source files before relying on it: Identify any design elements that could disqualify the SPCP safe harbor: is the program genuinely targeted at the disadvantaged class or does it risk excluding the intended beneficiaries?.
- Validate this output against source files before relying on it: Assess the interaction with fair lending risk: could the SPCP create reverse disparate impact claims from non-qualifying applicants?.
- Validate this output against source files before relying on it: Review the community needs assessment: does it satisfy the requirement for a documented determination of economic disadvantage?.
- 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 Special Purpose Credit Program Design, 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 regulation b 202.8 compliance assessment; design element risk analysis; reverse disparate impact risk evaluation; community needs assessment adequacy review. 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.

