COOKED
The core work — pulling credit bureau reports, verifying employment and income documents, comparing applicant data against fixed lending criteria, and approving or referring within a preset dollar limit — is exactly what automated decisioning engines have been eating since well before generative AI, and employment has already fallen sharply. What remains is exception handling and phone follow-up on incomplete files, which language models plus document OCR now do at usable quality. No license is required, the customer relationship belongs to the loan officer, and the credit policy is set by underwriting and risk, not by the clerk.
Part 2020 shock, part continued decline in the years since.
Median pay $40,100 → $50,080 -0.1% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
-6.2% 12,000 → 11,300 on the projections basis
Exposed, and shrinking
Both signals point the same way: the tasks are largely automatable and the BLS projects -6.2% by 2034. This is the case where the score and the forecast agree, and it is the one worth taking seriously.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~1,000 openings a year on average, including replacing people who leave.
CollectorInspectorAuthorizerControllerCredit ClerkInvestigatorCredit ExpertLoan VerifierTrust OfficerCredit AnalystCredit CheckerCredit OfficerLoan ProcessorCredit AdjusterCredit ReporterCredit ReviewerCredit VerifierBranch ProcessorCredit AssistantCredit AssociateCredit HistorianCredit ProcessorCall Out OperatorCharge Authorizer
Holding it up: judgment & accountability . Weakest point: embodiment .
Core tasks are already automatable Scorecard scoring, bureau pulls via Experian/Equifax API, DTI arithmetic, and stipulation checklists are rule-driven steps that FICO and Blaze decision engines execute at approval rates above 90% of file volume — the 3 rather than a 7 reflects that even your residual work, calling an employer to confirm a start date or re-keying a paystub, is now covered by OCR plus automated verification services like Truework.
Fully desk- and screen-based The job is a workstation, a headset, and a document queue; the only physical act is handling faxed or scanned paper, which the shift to e-signature and borrower upload portals has largely removed.
No licence, no signature requirement No NMLS registration, state license, or bonding attaches to a credit clerk — the lender's charter and the underwriter's signature carry ECOA and FCRA adverse-action exposure, and an approval you enter within your dollar authority is still the institution's decision, not yours.
Executes defined procedures on defined inputs You decide whether a thin file gets referred up or a stipulation counts as satisfied, which is real but bounded discretion — credit policy, cutoff scores, and approval limits are handed to you by risk management, and anything outside the matrix escalates rather than resolves at your desk.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (3/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (2/20) is whether the law requires a licensed human to sign. Trust premium (3/20) is whether buyers specifically pay for a person. Judgment and accountability (4/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 9 of this occupation's 13 points (69%).
Embodiment (1/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 28/100, still COOKED.
CFPB Circular 2022-03 and the 2023 follow-up already hold that ECOA/Reg B adverse action notices must give specific principal reasons even when a black-box model made the call. If examiners begin citing lenders for machine-generated reason codes that don't match the actual model drivers, banks staff a human reason-code validation step per denied file before the notice goes out.
Genuine two-tier structure exists here: the bureau-pull-and-match tier is gone, leaving synthetic-identity and first-party fraud files, small-business and self-employed income reconstruction from bank statements, and disputed-tradeline resolution. As the routine tier disappears entirely, the surviving headcount is concentrated in that tier — the score rises for the remaining jobs even as the count keeps falling.
A named-human-reviewer requirement for consequential credit decisions: Colorado SB 24-205 (AI Act, consequential decisions expressly include lending) obliges deployers to provide an opportunity to appeal an adverse AI decision 'with human review, if technically feasible'. If implementing rules or a successor state statute make that human reviewer a designated, auditable role inside the lender rather than a generic 'compliance function', the adverse-action reviewer becomes a defined job — the closest thing this occupation has to a sign-off seat. EU AI Act Art. 14 human-oversight duties for Annex III creditworthiness systems is the same mechanism for US banks with EU operations.
Fair-lending override logging: if OCC/FDIC exam guidance or a DOJ redlining consent order requires that every model override and every referred exception be documented with a named human rationale and retained for disparate-impact testing, the exception queue stops being clerical and becomes a decision-of-record with an owner.
The limit. Every lever raises quality of the surviving seats, not their number; headcount has already fallen sharply and none of these mechanisms creates demand. Trust premium has no plausible route — the borrower never meets this role and cannot pay for a human they don't know exists. Embodiment is fixed at floor.
| New York-Newark-Jersey City, NY-NJ | 810 | $59,370 +19% |
| Dallas-Fort Worth-Arlington, TX | 770 | $62,930 +26% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 410 | $63,400 +27% |
| Atlanta-Sandy Springs-Roswell, GA | 370 | $37,590 -25% |
| Los Angeles-Long Beach-Anaheim, CA | 290 | $52,530 +5% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 250 | $38,140 -24% |
| Boston-Cambridge-Newton, MA-NH | 230 | $55,360 +11% |
| Denver-Aurora-Centennial, CO | 190 | $50,820 +1% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 410 | $63,400 +27% |
| Omaha, NE-IA | 80 | $63,220 +26% |
| Dallas-Fort Worth-Arlington, TX | 770 | $62,930 +26% |
We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.
That is worth saying out loud next to a score of 13. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.