AI Isn't the Goal. Better Underwriting Is.

AI in insurancehas become a popular topic.

At a conference I attended recently, I heard from a number of well know reinsurers who are trying to implement AI in their business. What struck me wasn't the technology itself, but how success was being measured.

The conversation was dominated by questions like:

• How many AI projects are underway?
• How many people are using AI?
• How many tokens are being consumed?

Apparently, token usage has become a metric in its own right.

But none of these measures answer the question that really matters.

Is AI helping underwriters do a better job?

Are they making better risk decisions? Handling more business? Responding faster to brokers? Improving consistency? Delivering better outcomes?

Those are the measures that should matter.

Instead, many AI initiatives are still sitting outside day-to-day operations, experimenting at the edges of the business while trying to demonstrate success through adoption statistics rather than operational improvements.

The situation reminds me of Peter Thiel’s saying, “We asked for a flying car, instead we got 140 characters”. To rephrase it, insurers asked for AI, instead their staff got told to just use ChatGPT more. To be fair, insurers are not alone – there are stories doing the rounds of tech companies proudly stating how many tokens they used as a badge of honour.

What underwriters actually want 

In my opinion, we are asking the wrong question. We should be asking, "What do our staff actually need AI to do?"  Having worked alongside underwriters for more than 25 years, I've learned they aren't looking for another AI tool to experiment with.

• They're looking for help doing their jobs.
• They want answers to their questions without searching through multiple systems.
• They want key information available when they need it.
• They want repetitive manual work automated.
• They want to explain pricing decisions without spending hours gathering evidence.

Staff don't need another chatbot. They need an AI assistant to help them deliver services. And that means we need to be thinking operationally about AI, and how to embed it in our existing systems.

When you think about underwriting this way, the AI opportunities become much clearer. These are exactly the conversations we're having with clients every day.

They ask us:

• How can I find the right information for a quote across multiple systems?
• How can I fill in data for my pricing and risk models quicker?
• How do I compare similar historical risks?
• Please summarise a quote and highlight anything unusual.
• How do I manage my underwriter resources?
• How do I monitor my underwriting departments during busy renewal seasons?

In other words, AI needs to become a colleague to underwriters that can help them perform their work, not an ego project to win awards.

AI should work inside underwriting, not alongside it

That means answering questions using live underwriting data, completing operational tasks, supporting pricing decisions, summarising risks, helping managers balance workloads and allowing people to interact naturally with their underwriting platform.

In other words, AI should become part of the underwriting team, not another standalone AI experiment.

The winners in insurance won't have the smartest AI or the smartest implementations of it. They'll have AI that can safely execute real underwriting tasks within existing workflows and help the underwriters deal with more quotes and provide more insight. I.e. it will help them do their jobs.

The foundation matters

Of course, none of this works without the right digital foundation. AI is only as effective as the systems and data it can access. Insurers still relying on fragmented processes or spreadsheet-driven underwriting will struggle to unlock the full value of AI because the information AI needs simply isn't available in a structured, governed way. Before asking how to deploy AI, insurers should first ask whether their underwriting platform is ready for it.

The industry is moving quickly. Those with digitised pricing and underwriting platforms will be able to embed AI safely into everyday operations and realise measurable business value.

Those without them risk being left behind.

The future of underwriting isn't AI replacing underwriters. It's about giving every underwriter a highly capable digital assistant that helps them do more.