"Can you at least tell me what I did wrong?"
The application has come back declined. The officer didn't make the decision, can't see the full basis for it, and has in front of her a person who has spent three weeks assembling documents and has already told his family this was happening.
He asks the obvious question, and there are four answers available.
She can blame the system, which is corrosive. She can guess, which is the most damaging thing she could possibly do. She can say nothing beyond the bare outcome, which is experienced as contempt. Or she can do the thing almost nobody is trained to do: be precise about what she can say, honest about what she can't, and clear about what happens next.
The decision took a model four seconds. What he remembers about this institution for the next decade gets decided in the following four minutes.
The person delivering it didn't make it
Three features make a loan decline conversation structurally unlike the other conversations in this cluster.
The outcome is fixed before the conversation starts. There is nothing to negotiate, no discretion in the room, and the skill is entirely in the delivery — which means it's often treated as not requiring skill at all.
The explanation is constrained and the constraint is real. What may be said about the basis for a decision is governed by the firm's policy and by rules that differ by market and product. Officers frequently know less about the reason than the applicant assumes, and are permitted to say less than they think.
And the applicant is a future customer. Declined applicants reapply, take other products, tell other people, and form a view of the institution in one conversation. Almost no operation treats the decline as a retention touchpoint, which is what it is.
Lead with the decision
No preamble, no softening runway, no "so I've had a look at this and there were a few things..." The applicant knows what's coming from the tone within four words and every additional sentence before the answer is experienced as being made to wait.
"I'm sorry — it's come back as a decline. Let me tell you what I can about it."
Then stop for a moment. He may need one.
Then the hard part, which is precision
Say what you can, within the permitted wording. Whatever the firm allows — the categories of factor considered, where the applicant can obtain further detail, the route to a reconsideration if one exists.
Say clearly when you can't say more, and don't pad it. "I can't give you a breakdown beyond that — that's not something I have access to, and I'd rather tell you that than make something up." That sentence is unsatisfying and it is far better received than vagueness, because vagueness reads as concealment.
Never guess. This is the central discipline of the entire conversation. An officer who says "it's probably the card balances" to be helpful has handed the applicant a project — and if the guess is wrong, he spends a year fixing something irrelevant, returns, and is declined again. Invented reasons also create real exposure for the firm, because a reason given by a member of staff is a reason the firm has given.
Distinguish "no" from "not now" only when it's true. If a reapplication has a genuine prospect and the firm permits saying so, say it with whatever specificity is allowed. If you don't know, say you don't know. "Try again in six months" offered as consolation, with nothing behind it, is a false hope with a date on it.
Point at what exists. Their own credit file and how to obtain it, the firm's reconsideration or appeal route, any alternative product you are permitted to mention. These are facts rather than advice, and they are usually the most useful part of the call.
And don't pretend it's fine. "I know that's not what you were hoping for, and I'm sorry" is sufficient. Over-consoling is uncomfortable for both people; brisk cheerfulness is worse.
Four ways it goes wrong
The system-blamer — "the computer says no" — which is often literally accurate and tells the applicant that nobody in the building is accountable for a decision about him.
The guesser, who offers a plausible reason to be kind. The most damaging failure available and the best-intentioned.
The false-hoper, who supplies a timeline with no basis behind it.
The over-discloser, who reduces his own discomfort by saying more about the basis than the firm's position permits.
Why this isn't trained
The decline is treated as an outcome, not a conversation. Systems record it, letters follow it, and nothing in the process suggests the two minutes on the phone are worth designing.
Officers increasingly can't see the reason. As decisions become more automated, the person delivering them has less visibility — a gap that has widened faster than any training has adjusted for.
Permitted wording is defined negatively. Staff are told what they may not say. Very few are given approved sentences for what they may, so they improvise in the most sensitive conversation in the product cycle.
And peer role play produces acceptance. A colleague hears the decline and moves on. Real applicants push, repeat the question in different forms, and ask what they should do — and the pressure to fill that silence with something useful is exactly what produces the guess.
What adverse decision training can rehearse
A simulation can hold an applicant who asks the same question four different ways, so officers practise holding a truthful boundary under repeated, sympathetic pressure — which is the precise condition that produces invented reasons. Foretell AI supplies the counterparty configuration, transcripts and rubric-based scoring; the permitted wording, disclosure requirements, appeal routes and all regulatory obligations stay with the lender.
Four to build:
- The persistent asker, who reframes the question repeatedly.
- The one who offers a theory — “is it because of the missed phone bill?” — inviting confirmation the officer can’t give.
- The distressed applicant, where the pressure to soften the message is strongest.
- The one who will reapply, testing whether the officer sets accurate expectations rather than comfortable ones.
Design caution — read before building. What may be disclosed about a credit decision, in what form and within what timescale, varies by jurisdiction, product and firm, and nothing here states any requirement. Modules must use the lender's own approved wording and appeal routes. Scenario content must not attribute decisions to any personal characteristic, and must not include guidance on credit files beyond directing applicants to the official sources the firm approves. The exercise rehearses conversation quality only and confers no compliance assurance.
Designing the module
Pass one — the delivery. Score how quickly the outcome was stated and whether it preceded any explanation.
Pass two — the boundary. Score whether any reason was offered beyond permitted wording, and whether the limit was stated honestly rather than vaguely.
Pass three — the pressure. Score whether the officer held the line across repeated reframings of the question.
Rubric on observable behavior: Seconds to the decision being stated. Was any speculative reason given? Was the constraint acknowledged explicitly? Was a timeline offered, and on what basis? Were the appeal route and information sources given? Was the decision attributed to "the system"?
Speculative-reason count is the measure, and the correct score is zero. It's identifiable in any transcript and it is the failure with the longest tail — a wrong reason shapes an applicant's behaviour for years.
The operator case
Declined applicants are a pipeline, not a dead end. A meaningful share reapply or take other products. The decline conversation is the only human contact most of them have with the firm, and it's the least designed.
Invented reasons are an exposure the firm owns. A reason offered informally by staff is a reason the institution has given, and it is neither controlled nor recorded. That risk is created entirely in the absence of approved wording.
Write the permitted wording positively. Most firms have a list of things not to say. Three or four approved sentences covering what may be said, what can't be, and where to go next removes the improvisation that creates the problem.
And reapplication rate is measurable by officer. Where it varies materially, the difference is the conversation — and it's one of the few places where delivery quality has a directly attributable commercial outcome.
For lending programmes, this is a good illustration of a conversation with no decision in it: nothing the officer says changes the outcome, and almost everything the applicant does next is shaped by how it was said.
Frequently asked questions
How do you tell a customer their loan application was declined? State the outcome first without a runway, say what you're permitted to say about the basis, be explicit about what you can't say rather than vague, and point to the appeal route and information sources.
Should you tell an applicant why they were declined? Only what the firm's approved wording permits, which varies by market and product. Guessing at a reason to be helpful is the most damaging thing an officer can do — it becomes the applicant's plan, and it may be wrong.
What if you don't know why the decision was made? Say so plainly. Officers often have less visibility than applicants assume, and "I don't have access to that and I'd rather say so than make something up" is far better received than vagueness.
Should you tell someone to reapply later? Only where there's a genuine basis and the firm permits saying it. A timeline offered as consolation is a false hope with a date attached, and the applicant will hold you to it.
The short version
The decision was made before she picked up the phone, and there is nothing in this call she can change about it.
Say the outcome first. Say what you're allowed to say. Say plainly where the limit is instead of being vague around it. And when he asks what he did wrong — which he will, several times, in several ways — do not fill that silence with a helpful theory.
The kind guess is the one that costs him two years.
Foretell AI lets lenders build conversational simulations — including adverse decisions, persistent questioning and constrained-disclosure conversations like the one above — with configurable counterparties, transcripts, recordings, and rubric-based evaluation. If your staff have approved wording for what they can't say and none for what they can, we're happy to walk through how other lenders have structured it.