What is a decline reason?

A decline reason is the specific, stated reason a request was refused: a loan application, an insurance claim, a refund, a supplier's change of bank details. It tells the person what decided the outcome, in words they can act on. With AI agents, a good decline reason traces to the facts the agent used and the rule it applied.

Why do decline reasons matter?

Because a refusal without a reason leaves the customer with nowhere to go. They cannot correct a mistake, supply a missing document or complain effectively. A refusal with a specific reason is a form of service, an argument made in full in a refusal with a reason is good service.

In some markets the reason is also a legal requirement. In the United States, ECOA and Regulation B require creditors to give specific reasons when credit is denied, including when a model is used. "Did not meet our criteria" is not a reason. Training reviewers to meet that standard is covered in ECOA and Regulation B: training reviewers to give credit decline reasons with AI agents.

How does it work with an agent in the loop?

Decisions with significant effects on people, such as declining credit or a claim, are usually not taken by the agent alone. The agent reads the file, applies the policy, and drafts the outcome and the reasons at depth 2. A person with the authority to decline reviews the facts, checks that each reason is true and is the real deciding factor, and sends.

That review has a specific job. A reviewer must check three things:

1. The reason is accurate. The fact cited is in the file, from a source the business can stand behind.
2. The reason is decisive. It actually caused the decline, rather than being one of several minor points.
3. The reason is clear. A customer could understand it without knowing the policy's internal language.

The practical steps for credit are in how to draft credit decline reasons with an AI agent.

An example

Take a hypothetical consumer lender. The agent assesses an application and drafts a decline with two reasons: debt payments are high relative to income, and the applicant has a short credit history. The reviewer checks the file. The income figure the agent used came from a declared salary, but a signed payslip shows a higher figure. With the correct income, the first reason falls away. The reviewer sends the application back for reassessment instead of declining it. Without that check, the customer would have received a wrong reason and a wrong outcome.

How is it different from a refusal?

In our vocabulary, a refusal is a verdict on the agent: the action is outside the agent's bounds, so it may not take it. A decline is a decision about the customer's request. An agent can be refused permission to send a decline, which is exactly what happens when a decline is held for a person. The two often appear together, which is why teams confuse them.

A decline reason also differs from a correction reason. A correction reason explains to the agent owner what the agent got wrong; a decline reason explains to the customer why their request failed. The first is internal and should be general; the second is external and must be specific.

Related terms

Held actions, refusals and corrections connect here. Whether an agent may decline at all depends on the sector; see can an AI agent decline a loan application and the guide to AI agent skills by role.

Where the Academy fits

Pryme Intelligence Academy's Operator track covers how a governed agent decides and how human review works, including checking the provenance of each fact behind a drafted reason. Module tests are set in the learner's own industry and role, so a credit reviewer practises on credit decisions.