IMA Financial Group, an insurance and financial services firm with more than $8 billion in annual premiums, has been developing AI capabilities for internal firm use. Even at that size, IMA operates as an independent broker, so the firm is concerned about
Digital Insurance spoke with Garrett Droege, senior vice president and director of innovation and digital risk at the company, about agentic AI and data necessary to support these operations.
This article is from a longer interview and edited for clarity.
How is agentic AI useful to insurers?
You listen to these startups talk about what they're building, and how transformative it will be. A lot of them have agentic services as a solution. Yet, when you ask, how do you plan on getting the policyholder data and policy details, they make a lot of assumptions about how the insurance industry operates – that we have those records available in a structured format. That's not the reality for the vast majority of the industry.
It will require access to the core data systems that are just not built for it. Unless you're starting a from-scratch agency or brokerage, and you're building a modern environment that's open architecture, you really can't take advantage of agentic yet, not on a broad scale. You could do simple things like call center-type agentic functions.
What do carriers or agents and brokers have to do with data systems to support agentic AI?
A very large percentage of the top 50 property and casualty insurers have mainframe computers in their environments. They might have actuarial tables that were built in the 1950s and '60s on a mainframe. You can't take that out of that computer and plug it into a modern system. So they're having to reverse engineer literally their entire core data systems to streamline workflows with things like AI.
Today, with AI, everyone is recognizing we have to do this if we're going to be competitive in the next five to 10 years. If you don't, your only option is to exit through acquisition, because you don't have a hope to compete with someone that is fully taking advantage of agentic functions. They can run 30 to 60% more profitable than you simply because they're not requiring as many people to do it. That's going to be a real factor in competition. The will to make the change and upgrade core data systems is finally here, where previously it wasn't okay.
How do insurers distinguish the quality of an agentic AI service?
You have a hard time even testing whether this will work without updating your core system or components that you can modernize, that you could then try and deploy an AI solution around. It is very difficult to assess which AI platforms actually have something unique, versus which ones are just basically wrappers on top of some of the public AI models.
You have a rush of startups that figured out, okay, we can build an insurance specific function on top of ChatGPT or Gemini or Claude or whichever model. Insurance brokers and carriers won't know the difference. They won't have the in-house talent to be able to train the model to do this particular function or workflow. It's not a sustainable business model to build a solution on top of a piece of technology that you don't control. The companies that we're working with, we have vetted extensively. We've had our IT leaders, including AI developers, actually get under the hood and look at the technology and make sure one, it's secure, and two, that it is actually doing something novel, rather than something that we could have engineered in house on one of the public models.
Do insurers have the right data to support what they want to do with agentic AI?
I don't think that they do. The most simple, top 10 service requests for commercial insurance are certificates of insurance. It's updating insurance – I bought a new vehicle. We bought a new company. We sold one of those. We have new employees.
Those could be handled by agentic AI, either through voice or through chatbot, whichever people are comfortable with. You should have an omnichannel approach to customer service. People want different relationships in how they are serviced.
None of those functions that agentic AI can do for insurance are available without access to the AMS [agency management system] platform. Being able to modify coverage, you must have an open system that allows that kind of AI to function. We get sold on these solutions that look incredible, and then the reality is they can only take you so far today, if they can't access the core data.
Do independent agents and brokers have the right resources to use this technology?
We're at that breaking point where it will become increasingly difficult for smaller and mid-sized independent agents to compete with firms that have a better client experience, that's faster, more efficient, more user friendly, and more helpful from a risk management standpoint. Agents can also learn about risk profiles and make suggestions, just like a human could. You're going to have a more knowledgeable platform that's more helpful. Clients will ultimately gravitate towards that.
Agentic AI and the better experience that can bring will lead to a mass exodus of clients leaving agents and brokers, just because the user experience doesn't meet today's expectations. The people that figure out how to create a better user experience, they will win. We're trying to be in that mix and invest in technology and humans that are interested in deploying and using that technology for a better client and associate experience.
Will insurtechs or AI companies make this possible for smaller firms?
A lot of AI insurtechs are recognizing now that they can't access the AMS, but they can do things like embed themselves in an Outlook file. If you had AI watching your inbox and helping you triage insurance applications, service requests, things like that -- that's how the vast majority of independent agents and brokers work anyway. Instead of operating in the AMS, they're going around it and embedding themselves in Outlook or Google Workspace, or whichever email server that you use.
That's one way they'll be able to keep up, but that can only take you so far, because it still requires a human to take the information that comes back into your inbox and put that into the AMS. All service requests hinge on the information existing in the AMS platform. Insurtech has largely been a Band-Aid for the biggest problem in the industry, which is core data systems. It remains to be seen how much longer those Band-Aids can hold before the tide gives way.
