Your Car Insurance Is Priced on the Wrong Data | Armin Kia, Driverly

What we talked about.

Sunny Ray sits down with Armin Kia, co-founder and CEO of Driverly, to examine why motor insurance pricing leans on yesterday's data. After seven years at Admiral, Kia saw that even a leading insurer prices largely on demographics such as age, vehicle and postcode, using models that barely change for years. He explains how Driverly set out as a B2C broker, then pivoted to a B2B intelligence layer after the market turned and investors favored a leaner model. Driverly combines three waves of data: digital data, phone-based driving data with no black box, and connected car integrations, all turned into scores insurers can test through a proof of concept. Kia shares candid lessons on the cost of launching in a regulated space, a six to nine month pivot where both business and personal runways ran low, and how his three co-founders make decisions. He closes with a ten year view of insurance that is more personalized, dynamic, connected and preventive.

A former Admiral insider explains why car insurance pricing runs on static demographics and how live driving and connected car data can change it.

The questions, and the answers.

What did you see at Admiral that made you think pricing was built on the wrong information?

Even at one of the best insurers, the data is largely demographic. It is based on who you look like: your age, your vehicle, your postcode. That feeds models that barely change for years. When I saw how much more data exists, I saw a win-win-win for customers and insurers.

Seven of your classmates invested in Driverly. What made that an easy yes?

It came down to relationships and credibility, plus three things investors look at early on. First, the team and the trust we had built. Second, the technology, since we use AI much more smartly than the market. Third, the size of the problem. Team, tech and problem size were the winning combination.

You started as a broker and ended up as the layer underneath. What forced that turn?

Timing was the first factor. By the time we got our FCA permissions, the market had turned, with the Ukraine war, Brexit and COVID compounding an underwriting problem. Second, investors preferred B2B. It is leaner, avoids regulated activity and thousands of end users, and scales with better financials. It was the best decision we made.

How did your first insurer say yes, and what closed it?

It was a fast-growing MGA in Spain, and we knew the CEO and head of pricing from our Admiral days. But the numbers closed it. We ran a proof of concept, scoring a fraction of their book. They checked whether our score predicted claim risk, saw the value and chose to pay for it.

What was the most expensive thing you believed about motor risk that turned out to be wrong?

I was naive about launching a B2C business to challenge the big players. The FCA expects you to be ready to operate when you apply, and approval takes about 12 months. Then you need underwriting capacity from risk-averse insurers. Doing something untested in that market is very hard, and I paid a relatively big price.

Every founder has a near-death moment. What was yours?

It was our pivot, and it lasted six to nine months, the most stressful time of my life. We had little money, no revenue and low personal finances, so two runways were running out. My co-founders took no pay for a long period. We tested products, landed a first customer, then raised investment.

Can you describe the intelligence layer you are building?

We focus on three waves of data. First, digital data from online platforms. Second, our driving app, which captures driving data from the phone with no black box. Third, connected cars through OEM integrations. Our AI models combine all three into scores insurers can test easily, so we hide the complexity.

What does buying car insurance look like in ten years if you are right?

I see four shifts. It will be more personalized, covering policies and not just pricing. It will be more dynamic rather than static. It will be more connected to many data sources. And it will be preventive rather than reactive, stopping accidents before they happen instead of paying after.