Underwriting was automated early. Rules engines and scoring models have approved routine applications for years, and generative AI now reads documents and drafts summaries too. So the question for underwriters is less whether automation arrives and more how much work is left once it has.
The short answer: AI is reducing the number of underwriters needed, but it is not removing the role. The US Bureau of Labor Statistics projects a 4 percent decline for insurance underwriters from 2025 to 2035 and names the reason: automated underwriting software lets workers process applications quickly, reducing the need for underwriters. The work left is the cases software should not decide alone. Projections are estimates, not guarantees.
How exposed underwriting is
The International Labour Organization scores occupations' tasks from 0 (cannot be automated with generative AI) to 1 (could be done without a human operator), and groups them into gradients from 1 to 4, with 4 the most exposed; below them sit minimal exposure and not exposed (ILO Working Paper 140, 2025).
| Occupation (ISCO-08) | ILO category | Mean task score (0 to 1) |
|---|---|---|
| Statistical, Finance and Insurance Clerks | Gradient 4 (highest) | 0.64 |
| Insurance Representatives | Gradient 3 | 0.53 |
| Valuers and Loss Assessors | Gradient 2 | 0.45 |
The ILO index does not list underwriters as a separate occupation, so these are the closest groups. The ILO concludes that transformation of jobs is the most likely impact of generative AI, but the BLS outlook shows that in underwriting, transformation already means fewer roles.
What the BLS projects
| Occupation | Projected change 2025 to 2035 | Openings a year |
|---|---|---|
| Insurance underwriters | Decline 4% | About 6,800 |
Even with a decline, openings arise each year as people retire or change jobs. The BLS lists among underwriters' typical duties both employing automated software to evaluate risk and examining suggestions from underwriting software systems, so working with the software is already part of the job description.
Underwriting tasks software is taking on
- Scoring routine applications against established standards.
- Pulling and checking data from applications and records.
- Summarizing long documents, such as reports and prior claims.
- Flagging missing information and inconsistencies.
- Drafting referral notes and decision letters.
What stays with the underwriter
- Complex and unusual risks: cases that do not fit the rules or the training data.
- Reviewing automated decisions: knowing when to overrule the system and being accountable for the result. See human in the loop and accountability.
- Checking: generative AI can produce confident but wrong statements, which NIST calls confabulation (NIST AI 600-1). Summaries need to be checked against the file. See checking AI output.
- Fairness and data: applications hold personal data, and NIST lists harmful bias among the risks of generative AI. See AI agents and personal data.
- Negotiation: setting terms with brokers and clients on large or specialist cover.
How the job is changing
Fewer people review routine files and more of the work is exceptions, oversight and complex risk. Underwriters who understand how the models decide, and can explain and defend a decision, are better placed than those who mainly process standard cases.
See also will AI replace insurance agents?, will AI replace actuaries?, will AI replace bankers? and will AI take my job?
Pay and how to get in
The US Bureau of Labor Statistics puts the median annual pay for insurance underwriters at $81,370 in May 2025, against $50,980 for all workers. The typical education needed to enter is: bachelor's degree. See how this job compares in AI exposure by pay and AI exposure by education level.
Questions and answers
Will AI replace underwriters?
Not entirely, but it is reducing demand. The BLS projects a 4 percent decline for insurance underwriters from 2025 to 2035 because automated underwriting software lets workers process applications quickly.
How many underwriter openings are expected?
The BLS projects about 6,800 openings a year, mostly from people retiring or changing jobs.
How exposed is underwriting to generative AI?
The ILO does not list underwriters separately. Close groups score 0.64 (statistical, finance and insurance clerks, gradient 4) and 0.53 (insurance representatives, gradient 3).
What underwriting work stays human?
Complex and unusual risks, reviewing and overruling automated decisions, checking AI summaries, watching for bias and negotiating terms.