Data and AI risks drive MSI to change product processes

• Evaluating aerial imagery has to be constant and ongoing
• Thorough data reviews must also catch AI drift
• Anthropic partnership speeds up product design

Increased risks from managing property data and the increased presence of AI in the insurance industry led Millennial Specialty Insurance (MSI) to change its product management. MSI offers personal lines, commercial lines and multi-family business coverage. Its parent company The Baldwin Group entered a partnership with Anthropic, in May. 

Andrew Dalton, chief product officer at MSI, spoke with Digital Insurance about how its MGA business is changing its handling of property data and making use of AI technology. Dalton manages 60 colleagues in his role, including product managers, actuaries and predictive modeling teams.

This article is excerpted from a longer interview and edited for clarity. 

How do you set goals for handling the risks with property data?

The way we operate to bring analytics and insights into managing our programs is we have people with deep expertise in those spaces. We have the analytics horsepower on my team, and we engage deeply with third-party data vendors, modeling firms and our own team of experts as well to partner across the organization.

What do you see as the challenges in managing or evaluating property data?

There's a lot of opportunity in what is becoming available. For example, there's been a lot of discussion about aerial imagery – being able to leverage it, evaluate which ones work well, which ones have some weaknesses or some opportunities. Bringing that forward in a proprietary way is ongoing work that just constantly has to happen. There's more upside to what is becoming available from an analytics and data standpoint.

What is that upside? What benefits are you seeing?

With increased computing power and increased sophistication of models, a good example would be what has been happening with the personal flood market. Ten years ago, that was largely considered an uninsurable risk, but with increased computing power, with new data and analytics coming in, that's becoming a segment, including within MSI, where private carriers and private capital are interested in playing. There's certainly challenges there, but what is becoming possible from a data and analytics and modeling standpoint is going to help the industry navigate some of those challenges.

How do you respond to AI drift creeping into property data and figure out if data patterns are real?

When we use new sources of data, where we have people with deep expertise in those spaces, we do a really thorough review of those data sources. Any strong analytics firm is constantly monitoring for AI drift, constantly testing for changes happening in those patterns. AI and increased computing power enable doing that more frequently and easily, so that you can pick up on those things faster than would have been possible five years ago.

For MGAs like MSI, what unique challenges does AI present?

AI's analytics and computing power is going to come into how MGAs and other entities evaluate risk. We're going to see a pattern of really accelerating analytics sophistication at MGAs in a way that accelerates the risk for adverse selection if you're not keeping up with that. AI is not going to be able to solve creating the data to underpin that. You need unique relationships with distribution partners and vendors to get data.

That's really where the competition is going to land because the raw modeling itself is going to become more of a commodity in the next few years. Everyone's going to have access to these tools that can do amazing modeling work and be working 24/7. You're still going to need strong modelers who understand the business problem and the context, and can evaluate the output. But having that tool is going to be a game changer.

What are MGAs looking for from AI technology?

The history of MGAs is all about innovation, moving really fast, finding niches in the market. I think about how we can use AI to accelerate launching new programs and enhancing our existing programs, and how we can use AI to more effectively manage the day-to-day of our programs. For example, we are working on the core of a new product by having a team of colleagues work together with Claude for a day, which probably would have taken one or two months in the past. I'm very focused on using AI across the product pipeline as a differentiator for MSI in the market.

How has using Claude under Baldwin Group's partnership with Anthropic helped MSI?

The team that I oversee, and all 60 colleagues, are able to work faster because of it. Things that were traditionally more manual efforts – every role has some amount of that work – you can now push through the pipeline much faster. We're seeing a lot of benefits across all aspects of the product management lifecycle powered by Claude. Being able to build a product in a day or two, as I mentioned, would not have been possible without having Claude as a tool on people's desktops.