Adverse Action Notices & Compliant Decline Letters

When you decline an application, reduce a requested amount, or change terms unfavorably, the law requires a clear, accurate, and timely notice. The adverse action notice under the Equal Credit Opportunity Act (ECOA/Reg B) and the Fair Credit Reporting Act (FCRA) is one of the most compliance-sensitive documents a loan officer touches. AI can help you produce clear, consistent, well-written notices — but this is also the area where human verification matters most.

What You'll Learn

• What an adverse action notice must contain and why it's high-risk
• How to use AI to draft clear, plain-language decline communications
• How to keep AI from introducing fair-lending or accuracy problems
• The mandatory review steps before any notice goes out

What an Adverse Action Notice Requires

Under Reg B and FCRA, an adverse action notice generally must include:

• A statement of the action taken.
• The specific principal reasons for the adverse action (not vague generalities like "you didn't meet our standards").
• The ECOA anti-discrimination notice.
• If a credit score or credit report was used: the score, key factors, the credit bureau's contact information, and FCRA-required disclosures.
• Timing requirements (generally 30 days for completed applications, with specific rules).

These requirements are precise and consequential. A notice with a wrong reason code, a missing disclosure, or vague language is a regulatory finding waiting to happen. This is why AI's role here is drafting and clarity — never determining the reasons or guaranteeing compliance.

Where AI Helps: Plain-Language Clarity

The reasons for an adverse action come from your underwriting decision and your institution's approved reason codes — not from AI. But once you have determined the reasons, AI can make the notice clearer and more humane:

I need to write the explanatory portion of an adverse action notice.
The principal reasons for the decline (already determined by underwriting)
are: [LIST APPROVED REASON CODES, e.g., "insufficient debt-service
coverage," "length of credit history too short"]. Write a clear,
respectful, plain-language explanation of these reasons that a borrower
with no finance background can understand. Do NOT add, remove, or
reinterpret any reason. Do NOT include any rate, score, or legal
disclosure language — I will insert required disclosures from our
approved template. Under 150 words. Professional and humane tone.

Notice the heavy guardrails: AI explains only the reasons you supply, adds nothing, and stays away from the disclosure language that must come from your compliant template.

Softening Tone Without Changing Substance

A decline is hard news. AI helps you deliver it with dignity while keeping the facts intact:

Rewrite the explanation below so it is respectful and non-judgmental,
keeping every reason exactly as stated and adding nothing new. Where
appropriate, note that the applicant may reapply in the future if
circumstances change, without making any promise about a future
outcome. [PASTE EXPLANATION]

The phrase "without making any promise about a future outcome" matters — you never want a decline letter to imply a future approval.

Helping the Borrower Understand Next Steps

Within compliance limits, you can use AI to prepare general, non-promissory guidance you give verbally or in approved follow-up:

A borrower was declined for a personal loan due to high debt-to-income.
Give me three general, non-promissory talking points I could share if
they ask what they could do differently — framed as general financial
education, not advice specific to a future application or a promise of
approval.

This keeps you helpful and human without crossing into promises or steering.

The Fair-Lending Tripwire

Adverse action is where fair-lending risk concentrates. Three rules:

1. Reasons come from underwriting, not AI. Never ask AI to decide why an application was declined or to generate reasons. It must only restate the reasons your process produced, using your approved codes.
2. Consistency, not improvisation. Similar declines should produce similar notices. AI's consistency is an asset — but every notice must still match the actual file. Don't reuse a reason that doesn't apply just because it sounds good.
3. No protected-class language, ever. Reasons must relate only to legitimate creditworthiness factors. Have AI flag, but never insert, anything that could reference a prohibited basis.

The Mandatory Review Checklist

Before any adverse action notice leaves your institution, a human must confirm:

• The reasons match the actual underwriting decision and approved reason codes.
• All FCRA/Reg B required disclosures are present (from your compliant template, not AI).
• Timing requirements are met.
• No rate, score, or figure was invented or altered by AI.
• The language contains no promise of future approval and no prohibited-basis reference.
• Compliance has approved the template you're using.

AI can make decline letters clearer, kinder, and more consistent. It cannot make them compliant — that is your responsibility, and the responsibility never moves.

Key Takeaways

• Adverse action notices are highly regulated under ECOA/Reg B and FCRA; accuracy and required disclosures are non-negotiable.
• AI's role is plain-language clarity and humane tone — it must never determine or generate the reasons for the decline.
• Heavily guard prompts: AI explains only the approved reasons you supply, invents no figures, and stays away from disclosure language.
• Every notice gets a mandatory human review against an approved, compliant template before it goes out.