{ "source": "Cooked Index — occupational AI risk register", "page": "https://cookedindex.com/jobs/loan-officers/", "methodology": "https://cookedindex.com/methodology", "notice": "Verdicts are re-examined as evidence accumulates. Re-fetch before relying on this; the page above always carries the current score.", "scored_at": "2026-08-11", "model": "claude-opus-5", "occupation": { "title": "Loan Officers", "soc_code": "13-2072", "category": "Business", "us_employment": 274330, "median_annual_wage": 76690 }, "verdict": "EXPOSED", "risk_resistance": 37, "contested": false, "near_boundary": false, "dimensions": { "task_resistance": 6, "embodiment": 3, "liability_shield": 9, "trust_premium": 11, "judgment_accountability": 8 }, "reasoning": { "task_resistance": "Pulling tax transcripts, verifying W-2 income, ordering appraisals and title, and clearing DU/LP conditions are already handled end-to-end by Encompass/Blend workflows and OCR income parsers — a 6 rather than a 2 because commercial credit memos, self-employed cash-flow reconstruction from K-1s, and exception write-ups still get typed by a human.", "embodiment": "The job is a desk, two monitors, a phone and a DocuSign queue; the 3 rather than 0 covers the branch walk-ins, realtor open houses, and the occasional site visit a commercial lender makes to look at the collateral before funding.", "liability_shield": "NMLS licensing under the SAFE Act means residential originators carry an individual license number that appears on every disclosure and can be revoked, but the loan is approved by underwriting and the note is held by the institution — repurchase risk and TILA/RESPA exposure land on the lender, not on you, which caps this at 9 instead of the 14+ an attorney or appraiser carries.", "trust_premium": "Realtor and CPA referral pipelines are genuinely personal — borrowers pick you because their agent vouched for you and you answered at 9pm on a Sunday — but rate-shopping on Bankrate and refinance churn mean most consumers never call the same officer twice, which holds this at 11 rather than the 15+ of a private banker with a captive book.", "judgment_accountability": "Real discretion exists in structuring a deal to fit guidelines, deciding which compensating factors to document, and whether to push an exception up to credit committee, but the credit box, DTI ceilings, and investor overlays are written by someone else — you argue a file, you do not own the approval, which is an 8 not a 14." }, "rationale": "The modal loan officer spends most of the day collecting documents, verifying income and credit data, running applications through automated underwriting engines, and explaining terms — and algorithmic scoring plus document-extraction AI already does the analytic core better and faster. What holds is the origination relationship: NMLS-licensed mortgage originators and commercial lenders who source borrowers, structure deals that don't fit the box, and handle exceptions, appeals, and distressed files. Consumer/retail processing roles shrink hardest; relationship-driven commercial and jumbo/self-employed mortgage lending persists longest.", "outlook": "Expect meaningful headcount decline in consumer and processing-heavy roles by the mid-2030s, with surviving loan officers concentrated in commercial, complex-borrower, and referral-network origination.", "what_would_raise_it": { "levers": [ { "dimension": "liability_shield", "change": "Adverse-action and AI-model accountability rules that name a natural person: CFPB's 2023 circular already requires specific accurate reasons for ECOA/Reg B denials generated by complex models, and NMLS/state rules tie originations to an individual licensee's unique identifier. A state or GSE requirement that a named NMLS-licensed originator personally attest to the accuracy of AI-extracted income/asset data and to the adverse-action reason codes — as Fannie/Freddie reps-and-warrants already push back on lenders — would make the signature non-delegable rather than nominal.", "plausibility": "plausible", "would_add": 4 }, { "dimension": "judgment_accountability", "change": "Task-mix shift plus exception ownership: as automated underwriting clears the conforming file, what remains is credit-committee memos, guideline exceptions, non-QM/self-employed cashflow structuring, and loss-mitigation decisions. If bank regulators (OCC/Fed model-risk guidance SR 11-7 extended to origination AI) require a documented human credit judgment overriding or accepting model output, ownership of the ambiguous call concentrates in the officer.", "plausibility": "already happening", "would_add": 4 }, { "dimension": "task_resistance", "change": "Genuine two-tier occupation: routine document collection and conforming submission automates away, leaving deal structuring for entities/trusts/self-employed borrowers, CRE rent-roll and sponsor analysis, and distressed workouts. The score rises arithmetically as the routine tier exits the day, without any new law.", "plausibility": "already happening", "would_add": 4 }, { "dimension": "trust_premium", "change": "Narrow route only: commercial/CRE and jumbo borrowers already select lenders on the individual banker's relationship and ability to get a deal through committee, and brokers compete on that basis. If purchase-market realtor referral networks and small-business borrowers continue routing through named individuals rather than digital-only channels (Rocket/Better's share stalling in purchase vs refi is the observable signal), the premium holds and modestly rises — but for retail refi and consumer lending there is no such route.", "plausibility": "plausible", "would_add": 2 } ], "ceiling_note": "Even with every lever, this caps in the low-to-mid 50s and the gain is concentrated in commercial/jumbo/non-QM origination. High-volume consumer and conforming retail processing has no realistic lever on any dimension — no license attaches meaningfully, buyers shop rate not person, and the analytic core is already automated. Headcount can fall sharply while the surviving role's score rises.\n" }, "adjudication": null, "employment_history": { "points": [ { "y": 2017, "emp": 307240, "wage": 64660 }, { "y": 2018, "emp": 304950, "wage": 63040 }, { "y": 2019, "emp": 308370, "wage": 63270 }, { "y": 2020, "emp": 308700, "wage": 63960 }, { "y": 2021, "emp": 340170, "wage": 63380 }, { "y": 2022, "emp": 345550, "wage": 65740 }, { "y": 2023, "emp": 321090, "wage": 69990 }, { "y": 2024, "emp": 290530, "wage": 74180 }, { "y": 2025, "emp": 274330, "wage": 76690 } ], "from": 2017, "to": 2025, "change_pct": -10.7, "comparable_from": 2019, "spans_soc_revision": true }, "pivots": [ { "slug": "securities-commodities-and-financial-services-sales-agents", "title": "Securities, Commodities, and Financial Services Sales Agents", "verdict": "EXPOSED", "risk_resistance": 51, "median_wage": 78660, "overlap": 84, "skills_to_close": [ "Management of Financial Resources", "Systems Evaluation", "Monitoring", "Systems Analysis" ] }, { "slug": "financial-managers", "title": "Financial Managers", "verdict": "EXPOSED", "risk_resistance": 48, "median_wage": 166570, "overlap": 82, "skills_to_close": [ "Management of Financial Resources", "Systems Evaluation", "Monitoring", "Systems Analysis" ] }, { "slug": "personal-financial-advisors", "title": "Personal Financial Advisors", "verdict": "EXPOSED", "risk_resistance": 52, "median_wage": 105070, "overlap": 81, "skills_to_close": [ "Systems Evaluation", "Management of Financial Resources", "Systems Analysis", "Operations Analysis" ] } ], "license": "https://cookedindex.com/terms" }