It feels like you can't have a conversation about insurance these days without someone bringing up AI. And for good reason! The promise is huge: faster claims, smarter underwriting, happier customers. We're all excited about it.
In fact, recent numbers show that something like two-thirds of insurers—around 67%—are already dipping their toes in the water, experimenting with AI in some form. That's a lot of us. But here’s the rub, and it's a big one: most of these exciting projects never make it out of the lab. They're stuck in what I like to call "pilot purgatory."
They work great in a controlled, isolated test, but when it comes time to roll them out across the company? They hit a wall. And it’s not because the AI isn't smart enough. It's because of a much less glamorous problem that we absolutely have to talk about.
So, What's Really Holding Us Back?
Let's be honest. The real problem is our plumbing.
Think of your insurance company like a house. You've got your sturdy, reliable old appliances—the furnace (your core policy admin system), the water heater (your claims system), and the fuse box (your data warehouse). They've been working for decades. They're not pretty, but they get the job done.
Now, you go out and buy a fancy new smart home gadget, let's say an AI-powered assistant that can manage everything. You're excited to plug it in. But then you realize it has a completely different plug, needs a special kind of Wi-Fi, and can't even figure out how to talk to your 20-year-old furnace.
That, in a nutshell, is the problem we're facing. The fancy new AI tools can't talk to our old, reliable systems. This communication breakdown has a technical name: a lack of interoperability. And it’s the single biggest reason most AI initiatives are failing to scale.
The "Pilot Purgatory" Problem is Real
It’s deceptively easy to get an AI pilot off the ground. You can fence off a small area, feed it some clean data, and show off a cool demo. "Look! Our AI can read this one type of claim form and pull out the key information!" Everyone gets excited, pats each other on the back, and talks about the future.
But what happens when you try to connect that AI to your actual claims system? The one that's been patched together over 15 years and has a dozen different workarounds?
Suddenly, the AI needs to:
- Log into the legacy system.
- Understand a messy, inconsistent user interface.
- Pull data from three different screens.
- Connect to an external database to verify policy information.
- Push the final decision back into the system without breaking anything.
The cool demo falls apart. The project stalls. And everyone quietly moves on to the next shiny object, leaving another promising AI tool to gather dust on a server somewhere.
It's a Language Barrier, Pure and Simple
Another way to think about this is like hiring a brilliant, world-class expert who only speaks Swahili and dropping them into an office where everyone else only speaks English.
Your expert (the AI agent) is incredibly capable. They could probably solve your biggest problems. But they can't understand the instructions they're being given, and they can't communicate their findings. To get anything done, you'd need a human translator to sit next to them all day, painstakingly relaying every single piece of information back and forth.
That's what we're doing in insurance. We're trying to manually build these "translators"—clunky, custom integrations—for every single AI tool we want to use. It’s slow, it’s expensive, and it’s an absolute nightmare to maintain. It just doesn't scale.
Why Can't We Just Build a Bridge?
This is where the real challenge lies. Our core systems were built for a different era. They were designed to be stable, self-contained fortresses, not open platforms ready to connect with a dozen new technologies.
Getting them to "play nice" with modern AI requires a ton of work. You're dealing with ancient code, siloed data, and complex business rules that are often poorly documented. Building a custom bridge for every single AI agent is like building a brand-new road for every car that needs to cross a river. It’s just not practical.
We need a better way. We don't need a thousand tiny, rickety bridges. We need a superhighway.
So, How Do We Get Out of This Mess?
The answer isn't to throw out all our old systems and start from scratch. For most of us, that's a non-starter. It’s too risky, too expensive, and too disruptive.
The real solution is to build a foundational layer that sits between our old systems and our new AI tools. Think of it as a universal adapter or a master translator for the entire organization.
This "interoperability layer" would be responsible for one thing: communication. It would know how to talk to your old claims system, your modern CRM, third-party data providers, and any AI agent you want to plug in.
When you have this in place, everything changes. Your AI doesn't need to learn the specific quirks of your 15-year-old policy system. It just needs to talk to the universal adapter, which handles all the messy translation work in the background. Suddenly, you can plug in a new AI tool in days, not months. You can swap out one AI for a better one without rebuilding everything.
This is how we finally break out of pilot purgatory. It’s not about finding the one "perfect" AI. It’s about building the flexible, reliable plumbing that allows any AI to connect to our business and start delivering real value, at scale. It's a shift from one-off projects to building a true, future-proof capability. And honestly, that's a lot more exciting.
