LO
Get told when a verified record lands on this occupation → · Mark which of these tasks are yours (VOLO Pro, free) → · For a team or a graduating class →
Loan officer / credit officer
Decides, or recommends, whether money is lent: gathers the case, argues the exception, and is the name on the file when it goes bad.
Count what share of last month's approvals matched the model's recommendation exactly. That share is the part of your work the model already did.
This is not a probability of losing your job. It combines how much of the role's task load is exposed to automation with how far adoption has actually gone — useful for comparing occupations on one consistent basis, and for nothing else.
Covers consumer and small-business lending where a person is still in the decision. Large corporate credit is a different job with a longer cycle and more negotiation, and pure processing roles are closer to the administrative pages here. Regulation differs enormously on whether an automated decision needs a human review, and that rule decides more about this job than the technology does.
What is actually changing#
The unit of analysis is the task, not the job title. A role is not replaced — its task mix shifts.
Each tile is one task. Its size is how much of the job it is; its colour is where the task is heading. Click a tile to see what the judgement does not establish.
Sixty years of automating this task did not remove the occupation, and that is the most useful fact on this page — but it did change who gets in and how many. Read a high exposure here as a statement about the ordinary case only: the policy defines what ordinary means, and everything outside it still arrives at a person.
The exception surviving is not the same as the exception being permitted. Many lenders have narrowed or removed override authority entirely, which deletes this task without automating it — and where that has happened the stated reason is consistency rather than cost. A task can be irreplaceable and still be abolished by policy. China's rule requires a re-review mechanism, not a person on every case, and the lender writes its own triggers.
This is the task that employs the junior half of a lending team, so automating it removes the entry route rather than the work — the judgement needed above it was learned by handling a few hundred files. Nobody has said what replaces that, and the consequence arrives as a shortage of experienced officers five years later.
A signature is cheaper to keep than a decision, and a bank can keep the name on the file while moving everything that produced it. Human review that consists of clicking approve on a model's output is human review in name only. European supervisors who looked inside 13 banks found that 'human in the loop' sometimes meant only explaining a decision when asked; from outside, no measurement distinguishes the two.
Is this your job? Say so and this page narrows to your share of it.
A job title is a bundle of tasks bought together, and no two people hold the same bundle. Nothing is sent anywhere — it stays in this browser.
Read all 4 tasks in full — direction, reasoning and limits →
Recent changes#
NAICS 522, credit intermediation and related activities, all employees, seasonally adjusted. Read the peak date before the decline: the high point is March 2021, which is twenty months before ChatGPT was released, so the best-supported explanation for the turn in this series is the end of the mortgage refinancing boom and the interest-rate cycle, not automation. This is an industry series covering everyone a lender employs, including branch and back-office staff, so it cannot isolate lending officers; and it cannot see brokers and originators who are not on a bank payroll.
Verifiable change in hiring, headcount, hours or job scope. Highest weight — but causal attribution still has to be argued, not assumed.
Supervisors from the ECB, De Nederlandsche Bank and the Bundesbank reporting workshops held between May and August 2025 with 13 supervised banks, ten of which used AI for credit scoring. Their annex says a human in the loop remains central: for credit scoring, AI models support decision-making by humans, except for small retail loan originations, which are often automated; that the interpretation of human in the loop is still varied, in some cases limited to explaining the credit decision to the customer on demand; and that an emerging practice is more human oversight for higher-risk, larger applications. The supervisors say the findings come from a small sample and should not be generalised to the sector.
An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.
A banking regulation for deposit-taking lenders in China covering personal consumer and business loans (credit-card overdrafts excluded). Article 21 requires lenders to separate review from approval, approve loans within delegated authority, and — where they approve automatically online — establish a human re-review mechanism as a supplement, set the conditions that trigger it, and stop automated approval if post-loan management finds it cannot identify risk effectively. Article 5 requires accountability for each step of the lending process. It requires a re-review mechanism and triggers that the lender designs; it does not require a person to look at every loan, and it does not say how many cases reach one.
Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.
An EU directive that member states had to transpose by 20 November 2025 and apply from 20 November 2026. Article 18(8) requires member states to ensure that, where a creditworthiness assessment involves automated processing of personal data, the consumer has the right to request and obtain from the creditor human intervention, consisting of the right to a clear explanation of the assessment, to express their own point of view, and to request a review of the assessment and of the decision on granting credit. The recitals add that a review need not lead to credit being granted. The right applies on request; mortgage-secured credit is covered by a different directive, and unsecured credit above EUR 100,000 is outside this one except in limited cases.
Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.
The statute read article by article on the Ministry of Government Legislation's own portal, in the version in force from 11 September 2026; the article itself was newly inserted on 4 February 2020, which is the date recorded here. Paragraph 1 lets an individual require a credit rating company, or a prescribed credit information provider or user, to say whether automated evaluation is used for personal credit rating and for the setting, maintenance and terms of prescribed financial transactions — and if it is, to disclose the result, its main criteria, and an outline of the base information used. Paragraph 2 goes further than disclosure: the person may submit information they consider favourable, and where the base information is inaccurate or out of date may require it to be corrected or deleted and the automated result to be recalculated. Read paragraph 3 with the rest or the record misstates the law: the firm may refuse any of this where another law requires it, where refusal is unavoidable to meet a legal duty, or where complying would make it difficult to establish or maintain the commercial relationship — the last of which is broad and is the kind of clause a retelling drops. Procedure is left to Presidential Decree. This binds firms operating in Korea and says nothing about lending anywhere else. What it establishes about work is indirect but real: a legislature wrote a recourse route on the assumption that this scoring is already done by machine.
Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.
What this means for you#
Lending is the clearest case on this site of a job whose core calculation was automated sixty years ago and which still exists — but the entry rung is file preparation, and that is the rung being removed now. Get to the exceptions and the committee as early as anyone will let you, because that is where the occupation actually lives.
Your override authority is the thing worth watching, and it is being narrowed for reasons of consistency rather than technology in many lenders. If the limit you can approve within has not moved in three years while volumes have, that is the change happening to your job, and it will not be announced as automation.
Your options#
Four directions, each with its real constraints and one thing you can test this week. Continuing as you are is a legitimate choice — it just has to be a chosen one.
Move to the lending the model cannot score
Self-employed, seasonal, asset-backed and small-business lending are where the file does not describe the borrower, and that is where a person is still the mechanism.
It is higher-risk lending, so the job carries more scrutiny and the bad years are personal.
Count what share of last month's approvals matched the model's recommendation exactly. That share is the part of your work the model already did.
Move to model oversight
Several markets now give borrowers a right to human review of an automated credit decision or require lenders to keep a human re-review mechanism, and someone competent has to carry that review out.
It is a second-line role, which usually means leaving customer contact and a commission structure behind.
Find out who signs off your scorecard's annual review. If nobody can tell you within two calls, that is the vacancy.
Common questions#
The scoring half has long been done by models — a Korean statute already assumes it is done by machine — and the occupation still exists, which is the most useful fact available about this question. That history changed the work inside the job, not whether the job is there. The current wave extends the same mechanism to document gathering, which is the rung that employs juniors. The exception and the signature stay, because an exception is by definition a case the model scored wrong.
No date. Compute one number from last month: the share of your approvals that matched the model's recommendation exactly. That is the part of your work already done before you saw it. Then check a second thing that is not about technology at all — whether your override limit has moved in three years. Where a lender narrows override authority for consistency, that deletes the task without automating it.
It protects the signature, which is not the same as protecting the hours. Several markets now require a named accountable person and a right to meaningful human review of an automated decision — real rules with real force. But a bank can keep the name on the file while moving everything that produced the decision, and a review that consists of clicking approve on a model's output is a review in name only. Nothing measurable from outside distinguishes the two.
No, and that is why this occupation is the most useful precedent on the site. Statistical credit scoring is long established in consumer lending — a Korean statute written in 2020 assumes it is already done by machine — and the loan officer is still here. What changed was the entry ladder, the headcount and the wage structure — the people who remained do exceptions and accountability rather than calculation. Anyone reasoning about a newly exposed occupation should look at this one first, because it has already run the whole cycle.
Studying towards this?
These majors lead here. Their pages break down which of their competencies transfer and what graduates typically lack.
What these judgements rest on#
4 of 4 task judgements on this page are backed by a verified event and 0 are platform inference, each labelled where it appears. Behind them sit 2 technology dimensions, a reconstructed trajectory since language models reached the public, and 5 verified events.
See which technologies, how it got here, and the method →
Where it sits in the official classification: skills, knowledge, related jobs →
Other roles in the same function#
A company divides its work into functions before it divides it into jobs. These sit in Finance alongside this one — a fact about org charts, not a judgement that they are similar or that they are changing in the same direction.
Accountant / Bookkeeper · Auditor · Financial analyst · Quantitative analyst · Actuary · Insurance underwriter · Financial adviser / financial planner · Tax preparer / tax agent · Payroll specialist / payroll clerk