Artificial Intelligence (AI) has already had a big impact on the insurance sector. From predicting risk factors, to settling claims admin, automated systems are getting faster and smarter. IE caught up with Erez Barak, Chief Technology Officer at Earnix Excelerate 2025, to find out where we are heading in the future.
IE: What then is Explainable AI, I heard that phrase earlier at the event today? Is it something that is going to be part of the regulatory process in future?
EB: Yes, explaining a data model means show me the values, features, how you price risk and more. I think regulators will want more detail on the use of AI, how you interpret data, set guardrails etc in future. You see this in the US where you file new products. It will become more mainstream and in some ways it has to be built-into the AI systems you are using.
IE: OK, another scenario. Let’s say I am a broker with legacy data; portals, emails, phone calls recorded etc. How can I dig into that archived data and then overlay public data sources like CCJs, addresses, vehicle MoT advisories and more. Can Earnix help me crossmatch all that legacy data with public streams?
EB: Our first discussion would be let’s have more background about what your aims are, sharpen the tools.
IE: Say I want to find the roadblocks on price, where people don’t renew or stop the quote process. Basically I want to see where the customers are pulling out from my broker brand, for whatever reason.
EB: In general, you can use an AI Agent. That’s perfect for this tool. We have an AI Studio so we can build an AI Agent that can reason, read all your data, reach out to third party sources, and get a set of instructions from you, so you can limit the scope of what it’s searching for. It’s also compliant, because that’s how Earnix builds the AI Agent.
You then have your own advisor. Based on the data you and the AI modelled you can look at where some feature is missing, or a pain point, and then try to resolve it. You chat with the AI Agent and create a sub-set. This won’t be a light bulb “Oh-my-god” moment, but rather an organic kind of improvement of your product lines over time.
IE: You then put out a better value product into the market, to change that drop-out rate. Go more premium, or at least offer than option? Also does your AI Agent have a limit to what it can accomplish within one data set?
EB: An AI Agent can take data and shape it for any product line, plug it into your workflow. It can be applied everywhere and learn as it works. You can choose a start and exit point, but in theory it can keep evolving. You could bring in new lines to your existing AI Agent, or build another Agent to handle RV, vans, motorcycles etc. So maybe that gives you more niche insights.
IE: Plus, there might be different compliance issues like licences, values, underwriting panel maybe?
EB: Yes, there are always regional differences, compliance on traffic laws, licences and so on.
IE: Will we get to a point where AI is making most of the key decisions on risks or claims?
EB: It’s been overstated in the recent past, that AI Agents will take over the whole world. No, not likely. It’s more likely that in the future refined models of how humans and AI work together. The social side of things will advance things more than anything.
Right now, AI can ingest every book ever written, look at risk, interact with you. In future we will look at what AI can do in a very specific, focused way – what do you want to do with it? It’s a reasoning revolution. Humans can decide what it reasons about, who it reasons with. I don’t think insurance brands aren’t looking to replace traditional human empathy with AI decision-making, it won’t work. In the future it’s the partnership between AI and humans which will help insurers make their systems and products better. That’s the way ahead.
Be the first to comment