AI and Insurance Underwriters: What's Actually Happening and What to Do

The honest assessment

Insurance underwriting is one of those professions where AI isn't sneaking in through the back door. It's marching through the front entrance with a brass band. The industry has been data-driven forever, and the data people just got the most powerful data tool in history.

Here's what AI can already do in underwriting. Assess risk on standard personal and commercial lines faster and more consistently than human underwriters. Process applications by extracting data from forms, matching it against underwriting criteria, and generating a decision. Analyse claims history, public records, and external data sources to build risk profiles. Price policies based on sophisticated models that consider hundreds of variables simultaneously. Detect fraud patterns. Generate policy documents. Some insurers have already fully automated their standard personal lines underwriting... car insurance, home insurance, simple commercial policies go through with no human involvement at all.

What AI can't do well is handle the unusual. The complex commercial risk that doesn't fit neatly into a category. The marine cargo policy for a route through politically unstable waters. The professional indemnity cover for a business doing something genuinely novel. The large property risk where the survey reveals issues that require judgement about severity and mitigation. These require an underwriter who understands the risk not just mathematically but contextually. Who can look at a proposal and think "this feels wrong" before they can articulate exactly why.

The honest picture is that standard lines underwriting is being automated now. Not in five years. Now. Lloyd's of London has invested heavily in AI-driven underwriting platforms. Swiss Re, Munich Re, and the major composite insurers are all deploying AI across their underwriting operations. The trajectory is clear. The question is how far up the complexity chain AI can climb, and how fast.

Your exposure level: High

High exposure. Underwriting is fundamentally a pattern recognition and risk assessment task performed on structured data. That description could have been written as an AI capability specification.

The personal lines market is furthest along. Automated underwriting for motor, home, and travel insurance is essentially standard practice now among major insurers. The commercial lines market is following. Standard commercial combined, employers' liability, and professional indemnity for common professions are increasingly automated. The specialty market (Lloyd's syndicates, complex commercial, reinsurance) is the last bastion of human-centric underwriting, but even there, AI is being used to augment decision-making.

What this means in practice is that the number of underwriting jobs is declining while the remaining roles require more expertise. Entry-level underwriting positions are disappearing because the work they did is automated. Mid-level underwriters are under pressure because AI assists with the analysis they used to perform. Senior underwriters with deep specialist knowledge are, for now, relatively secure. But "for now" is doing a lot of work in that sentence.

The insurers aren't being subtle about this. Zurich, Aviva, and AXA have all publicly discussed AI-driven efficiency in their underwriting operations. The reductions in underwriting headcount are usually framed as "natural attrition" and "role evolution" rather than redundancy, but the net effect is the same. Fewer people, more technology.

The 90-day action plan

1. This week: let AI assess a risk you know. Take a recent submission you underwrote. Feed the details into ChatGPT or Claude and ask it to assess the risk, identify key concerns, and suggest rating factors. Compare its analysis to yours. You'll find it handles the standard factors well and misses the nuanced ones. That comparison is your curriculum for the next 90 days.

2. Week two: automate your documentation. Use Copilot to generate policy wordings, endorsements, and risk reports from your underwriting notes. The documentation side of underwriting is time-consuming and largely formulaic. If AI handles the first draft and you review, you've reclaimed hours.

3. By day 30: deepen your market knowledge. Use Perplexity and Claude to stay ahead of emerging risks. Climate change impact on property portfolios. Cyber risk trends. Geopolitical developments affecting marine and political risk. The underwriter who can speak knowledgeably about emerging risks has conversations that AI can't.

4. By day 45: learn to work with AI models, not against them. Understand how your company's AI underwriting tools work. What data do they use? Where do they make mistakes? What kinds of risks do they handle well and where do they struggle? Being the person who understands the AI's blind spots makes you the person who catches the errors. That's genuine value.

5. By day 60: build broker relationships deliberately. AI processes submissions. It doesn't take brokers to lunch. In the London market and commercial insurance generally, relationships drive business. Brokers place risks with underwriters they trust. Double down on the relationship side of your role. Be the underwriter brokers call when they have something unusual.

6. By day 75: specialise in complexity. Pick an area where underwriting is genuinely difficult. D&O liability for technology companies. Construction all-risks for complex projects. Fine art and high-value property. Areas where the risk assessment requires deep knowledge and the pricing isn't commoditised. This is where human underwriters will be needed longest.

7. By day 90: position yourself as a technical authority. Write a short piece about an emerging risk or underwriting trend. Present it internally. Share it with key brokers. The underwriter who publishes thought leadership on complex risks is not the underwriter who gets automated. They're the underwriter who gets promoted.

AI tools you should be using this week

• Microsoft Copilot for Work — Invaluable for the Excel-heavy work of underwriting. Rate calculations, portfolio analysis, and data manipulation are all faster with Copilot. Also useful for drafting policy documents and correspondence in Word and Outlook.

• ChatGPT for Work — Use it for risk assessment brainstorming, generating underwriting reports, and drafting broker communications. Give it a risk description and ask it to identify factors you should consider. It's like having a very fast but somewhat naive junior underwriter.

• Claude for Work — Better for analysing longer documents like survey reports, financial statements, and complex submission packs. Paste in a full submission and ask it to summarise the key risk factors and flag concerns. Handles detail well.

• Perplexity for Research — Essential for staying current on emerging risks. Climate data, cyber threat intelligence, regulatory changes, industry trends. When a submission lands on your desk for a sector you're less familiar with, Perplexity gets you up to speed quickly.

What to say in meetings

In underwriting team meetings: "I've been using AI tools to streamline our documentation and initial risk assessment. It cuts processing time significantly on standard risks. I'd like to discuss how we redeploy that time toward complex risk analysis and broker relationship development." Frame AI as a tool that makes you better at the high-value work.

When brokers ask about AI in underwriting: "We use AI to handle the data processing faster. That means more time for the proper analysis and the conversation about what the data doesn't tell us. You still need an underwriter who understands your client's business." Brokers want to know there's still a person making the decision. Reassure them without dismissing the technology.

If management discusses headcount: "The standard lines are moving to automated underwriting. That's sensible. But complex commercial and specialty risks need human judgement. I'd like to be part of the conversation about how we structure the team to handle both." Be part of the solution, not part of the problem to be solved.

If the worst happens

If you're made redundant from an underwriting role, your analytical skills transfer broadly. Risk management, compliance, financial analysis, consulting, and product development all value the ability to assess risk, work with data, and make judgement calls under uncertainty. The insurance industry also has a persistent talent shortage in some areas, so lateral moves within insurance are often possible.

Natural adjacent moves: risk manager (corporate or consulting), insurance broker (the client-facing side), claims management, loss adjusting, compliance officer, or insurtech product specialist. The insurtech sector specifically is growing fast and needs people who understand both insurance and technology. If you can combine underwriting expertise with AI literacy, you're exactly the person these companies are looking for.

One observation from having watched this industry for years. Insurance is cyclical. Underwriting headcount tends to contract during soft markets and expand during hard markets. AI is a structural shift layered on top of that cycle, but the cycle still exists. The underwriters who get through the lean periods are the ones with specialist knowledge, strong broker relationships, and technical credibility. AI tools help with all three, if you're willing to learn them. The ones who don't learn them are just hoping the cycle turns before the restructuring reaches their desk. Hope isn't a strategy. Never was.