ReviewDesk
← Back

ReviewDesk guide · By the Ingenious AI team · Published 14 November 2026 · Reviewed 14 November 2026

What does a good adverse-outcome letter actually look like? (Worked examples)

Generate a free adverse-outcome letter — no sign-up Build a free pre-use notice See pricing

The short answer. A good letter says the decision in its first lines, discloses the tool's role, lists principal reasons drawn from the person's own record with values they can check, and closes with the correction route and the human-review offer — 'a person, not the automated tool, will conduct that review'. The three examples below are fictional but structurally exact: they follow the five parts ReviewDesk's generator assembles for every letter, in Colorado, California and Illinois settings.

What is adverse-outcome letter examples?

An adverse-outcome letter is the written explanation given after an automated decision goes against a person. In ReviewDesk it is generated from the workflow record in five parts: the decision, the automated-tool disclosure, the principal reasons (entered per person, one per line), the jurisdiction note for Colorado, California or Illinois, and the review path — dated, numbered and filed in the audit trail, downloadable as a .txt.

Example one: a hiring decline in Colorado

The letter opens: 'We are writing about a decision affecting you. After screening for the warehouse associate role, we are not moving your application to interview.' Then the disclosure: an automated screening tool was used in connection with hiring, ranking applications on eligibility, availability for the shift patterns and a short recorded assessment. Then the reasons, numbered, in the applicant's own facts: '1. The availability recorded on your application covers one of the three shift patterns this role rotates. 2. Your recorded assessment scored below the benchmark for the numeracy section this role requires. 3. Two of the role's required certifications are not shown in the documents provided.'

The jurisdiction note follows in Colorado's frame — the explanation given within the regime's window, the correction path, the review right — and the letter closes with the routes: 'If any of this information is wrong, you can ask us to correct it — for example, if your availability has changed or a certification is missing from your file. You can also ask for a human review of this decision using the form on our notice page. A person, not the automated tool, will conduct that review.' Every sentence is checkable by its reader; two of the three reasons point at facts the applicant can repair. That repair potential is not an accident — it is what per-person reasons are for, per the reasons guide.

Example two: a tenancy decline with a correctable record

The decision line: 'We are writing about a decision affecting you. We are not able to offer you the tenancy of the property you applied for on 4 May.' Disclosure: a screening service's automated assessment was used, weighing rental payment history, court records and income against rent. Reasons: '1. The screening file shows an eviction filing from March 2023, recorded without an outcome. 2. Verified income of $2,400 per month against rent of $950; our criteria ask for income of at least two and a half times the rent, which is $2,375 — your income meets this, and it was not a factor against you.' Note the discipline in the second reason: a factor that was tested and passed is said to have passed, because a reader told only negatives will assume every factor failed.

The correction paragraph does the letter's heaviest work: 'The filing at reason 1 is the record you may be able to correct. If the case was dismissed or decided in your favour, send us the court disposition and we will re-assess your application on the corrected record.' The review offer follows with the form route and the person's-review promise. This letter's real-world twin, in the housing guide, produced a verified disposition the next morning and an approval on day two — because the letter identified the exact record, its date and its missing piece. Specificity is not a style preference; it is the mechanism by which errors get fixed.

Example three: a credit decline in a California workflow

Decision first: 'We are writing about a decision affecting you. We cannot offer you the credit facility you applied for on 18 June.' Disclosure: an automated assessment banded the application using credit-file data, the application data and affordability inputs. Reasons: '1. The credit file we received shows two accounts in arrears in the last twelve months, one by more than ninety days. 2. Your current committed outgoings leave monthly headroom below the level the repayment schedule requires, on the income figures you provided. 3. The facility requested is above the maximum our criteria set for a first facility.' Three reasons, three different remedies: correct the file, revisit the arithmetic, or resize the ask — and the letter names all three routes implicitly by naming the reasons precisely.

The California note then states the workflow's arrangement accurately — the opt-out position for future applications and the human-appeal route for this decision, per the routes guide — because a California letter that reads like a Colorado letter misdescribes the reader's rights. Jurisdiction notes are generated from the workflow for exactly this reason: the reviewer's reasons are personal, the regime's frame is structural, and confusing the two authorships produces letters that are warm and wrong.

What all three examples share — the anatomy

Five parts, always in the same order, because readers learn the shape and can find themselves in any letter from the same sender. The decision, stated as a fact about the person, in the first lines. The disclosure, brief and unembarrassed. The reasons, numbered, per-person, with values and thresholds where the record holds them. The jurisdiction note, in the regime's own frame. The routes — correction, review, and the human promise — at full size, with handles. Around the five parts: the letter's number and date up top, the workflow's identity, and nothing else — no marketing, no cross-sell, no survey invitation, in line with the tone guide's provocations list.

The anatomy is also a diagnostic. Lay any of your current decline letters beside it and mark which parts exist. Most legacy letters contain part one (buried), a hint of part three (generic) and an accidental part five ('contact us if you disagree'). The generator's contribution is not eloquence; it is the impossibility of omitting a part. Structure is what quality looks like when it has to survive a tired Tuesday in February with a hundred letters to send.

What the examples deliberately do not contain

No scores. A model score appears in none of the three, because a score is not a reason a reader can act on. No blame vocabulary: nobody 'failed to demonstrate' anything; records show, criteria ask, gaps exist. No statistics about other applicants — the reader's case is the letter's whole world. No apology theatre and no counterfeit warmth; the register is respectful because it is useful, not because it performs feeling. And no legal fog: the jurisdiction note states rights as actions with routes, in the plain-language discipline of the writing guide, rather than as citations to be admired.

Also absent: any sentence the record could not support. If a reason cannot be traced to this person's decision record, it does not appear — the generator asks for reasons per applicant precisely so that the letter's facts have an author and a source. A letter is evidence about a decision; its sentences should be able to survive the question 'where is that in the file?' without exception. That standard, applied ruthlessly, is what makes the examples read as calmly as they do: there is nothing in them that needs defending, only things that need reading.

How to use these examples in your own programme

As a sampling standard. Print the anatomy, take five of your live letters monthly, and mark the parts — the exercise from the review checklist applied to correspondence. Score reasons for checkability: could this reader verify, correct or accept each one from the letter alone? Then generate a letter for a synthetic case at /try/letter with your own typical facts and compare shapes. The gap between your letters and the anatomy is your template backlog, ordered by whichever missing part your applicants feel first — usually the identifiable record in the reasons, and the real review route at the end.

Adopt the examples' one luxury: time. Each was generated the day its decision was made, while the record was open and the reasons were fresh. Letters written in monthly batches inherit batch thinking — generic reasons, stale routes, a tone set by volume. The generator removes the excuse for batching, because a letter costs minutes when the record is already assembled. Same-day letters are the examples' least visible and most replicable feature.

Your five actions for adverse-outcome letter examples

  1. Check every letter against the five parts: decision, disclosure, principal reasons, jurisdiction note, review path — in that order.
  2. Strike any sentence the person's own decision record cannot support, and any score offered as a reason.
  3. State factors that were tested and passed as passed; a reader told only negatives assumes everything failed.
  4. Give the correction route its target: the exact record, its date, and the document that would settle it.
  5. Generate on decision day, and sample five live letters monthly against this anatomy.

Worked example: the letter that produced its own correction in one day

The tenancy example above is drawn from the pattern in the housing guide's worked case: a fictional applicant, an eviction filing recorded without an outcome, a letter that named the record, its date and its missing disposition. The applicant's reply arrived the next morning with the court document attached; verification took one lookup; the re-assessment cleared the remaining factors; approval issued on day two. The landlord's total effort after sending was under an hour, most of it the reviewer's reading.

Set that beside the same landlord's previous form decline — 'your application was unsuccessful; our decision is based on information from a screening service' — which historically produced either silence (and a lost good tenant) or a solicitor's letter (and a month of correspondence). The new letter is not kinder in any sentimental sense. It is simply complete, and completeness turned a potential dispute into a same-week tenancy. That conversion — from argument to paperwork — is the entire commercial case for the five-part letter.

Frequently asked questions

What are the five parts of an adverse-outcome letter?

The decision; the automated-tool disclosure; the principal reasons, numbered and specific to the person; the jurisdiction note for Colorado, California or Illinois; and the review path — with the correction route stated and the promise that a person, not the tool, will conduct the review.

Why do the examples avoid quoting a model score?

A score is not actionable: the reader cannot verify, correct or accept it. The reasons translate the score back into the real-world factors — records, values, gaps — that the reader can act on.

Should a letter mention factors the person passed?

Yes, where a factor was materially tested. Saying income met the threshold and was not a factor against the applicant prevents the natural assumption that everything failed, and it makes the true reasons stand out cleanly.

Can I generate a letter in this shape now?

Yes — the generator at /try/letter assembles exactly this structure from the applicant, decision and reasons you enter, runs free with no account, and produces a downloadable letter.

Is this guide legal advice?

No. ReviewDesk is templates and workflow tooling, not legal advice. The examples are fictional and structural; consult counsel for your letters' legal content.

Keep reading: How do you respond to a human-review request, step by step? · What happens after a human review decides — for the applicant and for the business? · What should an annual ADMT programme review actually examine? · How do you train hiring managers to work inside an ADMT notice and review process? · How do you choose ADMT compliance software? (A buyer's checklist) · All ReviewDesk guides · ReviewDesk home · Try the free desk · Free letter generator · Free notice builder · Pricing

ReviewDesk is built by Ingenious AI. Templates and workflow tools — not legal advice.