We Need a New Way to Talk About AI in Insurance 

We’ve had more conversations about generative AI (gen AI) than I can count during the past year with CIOs, COOs, underwriters, claims leaders, data teams, you name it. The same questions keep coming up: Where do we start? What does “good” look like? How do we scale this responsibly? 

It’s exciting, but also deeply confusing. Because for all the noise around AI, the insurance industry still lacks a shared, standardised language to talk about maturity, risk, and readiness. Everyone’s piloting something, but few know where AI actually fits into a bigger picture. 

Sapiens tried to fix that by borrowing a concept from a totally different industry: autonomous vehicles. Back in 2014, when “self-driving” meant everything and nothing, SAE International, the leader in connecting and educating mobility professionals, introduced a five-level framework. It clarified what autonomy actually looked like, from manual control to full automation. This framework changed how engineers, regulators, and manufacturers worked together, giving them a common perspective and goal. 

We think insurance needs the same clarity. 

Our latest whitepaper applied that model to generative AI in insurance. It breaks down automation maturity into five levels, starting at the completely manual (Level 0), and ending with full AI-driven operations (Level 4). It’s not about overselling the tech, it’s aimed at helping decision-makers understand where they are today, what’s realistically next, and how to scale in a way that’s safe, explainable, and valuable. 

Let me be clear: this isn’t theory. We’re already seeing real progress. Claims triage in personal lines is already hitting Level 2. Some digital-native players are experimenting with Level 4 for specific lines, like renters or pet insurance. 

Most insurers are still at Level 1, which creates a huge competitive advantage for the minority of early adopters. The majority are doing gen AI pilots, like summarising call transcripts, extracting insights from documents, speeding up back-office tasks. That’s fine. In fact, it’s smart. What matters more, though, is having a roadmap to move forward strategically, instead of reactively. 

And here’s the part I can’t stress enough: the tech is only half the story. What separates the leaders in this industry isn’t how many AI tools they license, it’s how they prepare their data, empower their people, and embed governance. 

AI needs to be explainable. Outputs must be auditable. And the tools should be accessible enough that business users (not just IT) can work with them confidently. That’s why we’re building solutions that put power in the hands of underwriters, claims adjusters, and frontline staff. 

There’s no shortcut to Level 4…but there is a structured path. 

If you’re a CIO trying to modernise your stack, a COO under pressure to cut costs, or a Chief Data Officer thinking about model risk, then this framework will help you navigate the coming trade-offs. It’s already helping our clients benchmark progress and align teams on what real automation looks like. 

You can read the full whitepaper here. We created it to bring clarity to a fast-moving conversation, and we’d love to hear what you think.