AI is revolutionizing the way insurers deal with underwriting and claims management. However, adoption still faces barriers that go beyond implementation.
The most frequent blockers in adopting AI are not technological. Though insurers start AI projects with trusted vendors and a clear understanding of why the work is necessary, many stall. Our team has seen underwriters across commercial and specialty lines hesitate to rely on scores generated by AI. At the same time, claims teams worry about the possibility of using models to affect settlement decisions.
A Pilot That Stalled — And What Changed
I find it challenging to persuade underwriters to trust AI-generated recommendations. In a commercial P&C insurer pilot project, the AI model was ready in four weeks, but rollout stalled for several months because underwriters didn't trust scores without context. Adoption only took off after we explained how our AI advisor worked in real life. For this, we asked our partner to provide 4,000 historical data points, which we then used to train the AI model. Also, we did the following:
- Showed the top factors influencing AI scores
- Allowed underwriters to override AI outputs
- Offered to keep an audit trail of all recommendations and decisions
- Embedded AI results in the tools they already use
As a result, we've got a data-rich advisor that calculated triage, appetite, and winnability scores in a matter of minutes, but more importantly, a solution that underwriters trusted enough to start using. Such trust turned a pilot into a full-fledged software product, taking underwriting processing to a new level.
Transforming Manual Workflows Into Digital Journeys
In the case described, AI helped underwriters transform traditional, often outdated and manual, processes into an automated digital journey. Triage scores are calculated more accurately as the platform ensures that data is complete. Appetite matches submissions against preferred segments and considers the company's guidelines and rules. Winnability predicts the likelihood of winning the deal. All scores are calculated automatically, saving underwriters' time for final decision-making.
Overcoming Fears of AI Replacing Professionals
Another challenge is the fear that AI could replace underwriting and claims specialists. The key is to convince underwriters that AI is a helper rather than a rival. On a project that required a certain level of automation in claims, our solution was to integrate natural language processing to extract data from documents supporting claim submissions from customers. As a result, claim managers have 25% more time for complex cases requiring more attention and direct communication with clients.
Asking for feedback is also crucial. It allows you to discover when AI predictions and recommendations were right or wrong and use that information to refine the models. And when people see their feedback improve models, trust accelerates.
In measuring the impact of AI in underwriting and claims, it's not about providing ROI to leadership. It's more about building credibility, so the people who use AI believe it works.
From our experience, here's what works for measuring the impact of AI:
- We measured the current state before AI was introduced (average triage time, claim cycle time, loss ratios, etc.)
- Together with customers, we tracked the usage rate and override frequency
- Our experts looked for early wins during one quarter to scale further
Success doesn't mean integrating complex algorithms only. It comes from addressing AI adoption challenges, delivering measurable results, and building solutions that insurers trust.
