In this Peer to Pier edition of Insurance Business TV, Matteo Carbone, founder and director of IoT Insurance Observatory, makes a self-described contrarian case that AI-native customer journeys won't make a dent in agent and broker distribution over the next five to 10 years.
Carbone starts by separating hype from reality: insurers have used AI for decades, and today's real gains are in efficiency, from document management to speech-to-text. He then gives three reasons he is skeptical about AI agents replacing intermediaries: autonomous AI transactions are still an exotic idea even for everyday purchases, insurance is an infrequent, high-consequence purchase rather than an impulse buy, and policyholders value a human accountable for the decision. He adds that even if confidence arrived, the shift would take years because people change insurers so rarely.
[00:00] [music]
[00:07] [music]
[00:13] Paul Lucas: Hello everybody. Welcome to Insurance Business TV. We are here at Insurance Fest 2026 for another of our peer-to-peer interviews. Today I'm catching up with Matteo Carbone. He is founder and director of IoT Insurance Observatory. So Matteo, first of all, let's talk about AI. We can't escape it, of course, right now. What would you say is the most meaningful way that AI or connected data is perhaps already changing underwriting or claims versus maybe a few years ago and where is it all just still hype?
[00:42] Matteo Carbone: Let's first of all clarify that insurers have used AI for a couple of decades.
[00:52] I remember probably 15, 16 years ago Allianz in Italy was using already neural networks within the claim process to connect witnesses, policy holders, doctors, and body shops that was popping up in different claims to detect fraud.
[01:14] So, obviously over the years insurers have used more AI. More recently we have seen generative AI coming on the table of insurers. So, in many of the processes that are inside the kitchen of insurers today, AI has already helped to increase the efficiency. So, we are talking about document management. We are talking about text, speech to text. All things that can make more efficient pieces of the insurance process.
[01:54] Paul Lucas: Yeah. And tell us, the distribution side of things as well. Have you seen any misconceptions around the way that insurance distribution and AI go together?
[02:03] Matteo Carbone: Yeah, let's say that we have seen over the past 6 months a bit of craziness around the hypothesis that ChatGPT or AI native customer journeys can disrupt the traditional insurance distribution.
[02:26] So, about this, I think that many insurance executives over the months when asked what they think about it, defended the corporate business, the more complex product. Somehow, instead, they let the door open about a scenario where maybe this can happen on retail auto or homeowner insurance.
[02:58] So, I'm extremely skeptical on this. I can define myself contrarian. I believe that this kind of insurance distribution, looking at the global volume of insurance, will not make a dent. And even if we go on the detail of the retail policies, auto and homeowner insurance, in the next 5 to 10 years, this will not change the way in the different markets that are obviously with different distribution landscapes, policies are distributed.
[03:35] For at least three reasons. So, first, the idea of an AI agent doing transaction autonomously is still an exotic thought. It's not yet happening, even to purchase toilet paper.
[03:57] So, let's have a use case demonstrated. When people will be more confident in letting someone doing transaction autonomously for them on things that are not high impact on their life, we can start to talk about it and we are not yet there. But let's make an act of faith and say at some point the technology will be there. We can discuss when, but let's say at some point we will be there.
[04:29] Insurance is not a product that is based on an impulse purchase. It's not frequent. It's something that has potentially high consequences. So, we have seen that even initiatives over the past 10 years that have lowered the frictions have not changed the weight of the intermediated distribution, so what is being sold by agents and brokers.
[05:02] We have some markets as the UK, Australia that have been digital on auto and home and other insurance for 15, 20 years. But they are exceptions. There are some players that are doing extremely well on the direct channel as Progressive. But still independent agents in the US have the highest share of auto and home and other insurance.
[05:31] If you look at continental Europe, agents are dominating the market in Italy, France, Germany. So, let's think about these policyholders that with all the digital options they had available for the past 10 years have not chosen to move to a direct purchase because they feel that they prefer to have a safety net. A human accountable that help with them in the decision and if something wrong they have this safety net that says someone to blame.
[06:02] So, I don't see in the next 5 to 10 years these people, a significant number of these people, stopping to rely on the human and letting a machine not only do the automatic translation but also bearing this trust, representing the safety net.
[06:22] Then there is one last aspect that is about the time. So, even when a significant number of people feel confident, then from this moment to the moment when the weight of the different distribution channels change can be a pretty long time frame because only a small number of people change insurer frequently.
[06:50] I did a survey 2 years ago on home insurance and 60% of the US families own the same insurance coverage since they changed home the last time.
[07:07] So, even assuming that magically one day everyone is confident on the machine buying for them the policy, they will do it only when they have the need, when they change home. So, it will take a really long time. So, all these expectations of the disruption of the agents and broker business in 5 and 10 years looks silly.
[07:30] Paul Lucas: Okay, interesting. I'm sure a lot of agents and brokers will be quite relieved at what you've got to say there but if I can just focus on perhaps some areas where you think AI might make a big difference. So, where would those be?
[07:43] Matteo Carbone: Well, they mentioned it before about efficiencies on the processes. Clearly, it can have a significant impact on where labor force is allocated.
[08:03] So, if you automate tasks of the contact center, all the workforce that today is allocated on the customer contact to deal with a claim, this somehow is reshaped. Doesn't mean that insurer will choose not to have a human that at some point is available to talk with their policy holder, but the activities that this human will do will be different than today.
[08:25] So, it will be more talk with a client, less administrative activities. So, it makes totally sense that you will have a significant reduction in the number of FTEs that are necessary for this activity.
[08:42] We have started to see some early signs on assistance company in Europe, for example, that have the largest call centers because their business is based on frequent claims. So, the contact center is one of the key areas where they are allocating workforce. They have started to reduce the people allocated to this process.
