Insurance leaders are coming to a consensus: AI might actually change the way things get done. (Credit: Shutterstock)
Over the past five years, AI leaders in the insurance industry have created 6.1 times the total shareholder return of AI laggards. A staggering stat? Yes, and the lack of debate or shrieking doubters speaks even louder than the stat itself.
As McKinsey puts it: “Although few insurance companies are extracting meaningful value from AI across the full value chain at scale, best-in-class insurers are taking a domain-based approach to transformation. They choose certain business functions — such as distribution, pricing, underwriting, claims or investments — and comprehensively revamp how that function operates..."
"So far, domain-level rewiring with AI has had a measurable impact on key parts of insurance businesses, including a 10% to 20% improvement in new-agent success rates and sales conversion rates; a 10% to 15% increase in premium growth; a 20% to 40% reduction in costs to onboard new customers; and a 3% to 5% accuracy improvement in claims.”
Despite the early success and enormous energy around AI, there’s far more untapped growth ahead. Sprinkling a little AI on top of existing workflows only gets you so far, as McKinsey points out. The most meaningful value comes from rethinking entire functions, and a key obstacle is legacy technology.
The Old Core
The legacy systems that everyone knows, the ones that have been the way the insurance industry has worked for decades, were never built with AI in mind. Instead, they were built through M&A. The Old Core of eight companies have acquired no less than 90 companies over their 36 year average age.
What happens when you have to integrate all kinds of systems that were never designed to work with each other?
Compromises. Lots and lots of compromises.
AI can never bolt on neatly, because it can’t access all the data it needs.
And when systems aren’t AI-native, it hurts performance.
This is a problem across all industries, but it has hit the insurance industry particularly hard. Insurance is a full generation behind on technology, having missed out on the cloud.
The Old Core systems are showing their age: even now they’re far more like the web-based client-server of the early 2000s than the future-ready, AI-enabled systems they need now.
Insurance leaders, understandably, are frustrated. Why is everything that’s so easy in mobile and cloud and AI in their everyday life so impossible in their systems at work?
The answer is simple: their Old Core systems were never built with AI in mind. In theory, model context protocol (MCP) can work with something as old as a mainframe. But in practice, the libraries with hundreds of thousands of downloads and contributions are focused on how this protocol interacts with API-native technologies.
AI will work around the Old Core. But it can’t work within it.
The industry is looking for workarounds that let it maintain the Old Core while getting some of the benefits of AI.
Nobody will be surprised when the Old Core buys or builds AI solutions.
But we also shouldn’t be surprised when the results are disappointing compared to what we all know is possible. It’s an open secret that the “cloud” implementations of this Old Core software are lift-and-shift monoliths. Changes that should be easy take too long and cost too much.
Progress will depend on a never-ending series of workarounds and management’s willingness to fund them year after year. I’m not sure that’s sustainable. And even if it was, I’m positive that’s not smart.
Worse, workarounds assume that the future will resemble the past. It’s hard to predict what critical new capabilities AI could make possible next year, never mind over the next five years or the next ten. The only thing we can be certain of is that your speed to opportunity will be dictated by how well your data is organized. The more silos, the more incompatible systems, the more bolts and duct tape and baling wire holding the stack together, the slower your speed to opportunity is likely to be.
When the industry thinks about AI, we should think less about automation and more about driving better business outcomes. Reducing keystrokes, summarizing documents, and auto-filling forms are all useful, but hardly transformative.
Proactive portfolio management and better decisions across the full policy lifecycle are both possible. But it’s not as simple as bolting AI onto a legacy system. Just because something is technically possible doesn’t mean it’s going to work well at scale.
To paraphrase renowned executive coach Marshall Goldsmith, "what got the insurance industry here won't get it there".
Success or failure with AI depends on the approach a company takes.
Make sure your company is on the right track.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.