Path one: Build it yourself

I find that many organizations’ first instinct is to bring AI-tooling in-house. Often a technology leader or Chief Claims Officer is confident the team can bolt the technology onto the claims system they already run. 

I think the vision is right and makes sense on the surface: you already have the team and the system, so bolting on AI seems like it would allow you to simply optimize what you have while maintaining control. However, this isn’t always straightforward in practice.

The case for it: You own the process. You know what's been built, you shape the design, and you own how claims get handled. Every claims leader wants good outcomes, and that feels easiest to achieve when you are in the driver’s seat. 

The catch: It’s all too common to come up short on resources. A build like this takes far more engineering than most technology vendors assign to one client. About six to eight months in, many leaders come back to ask for a bigger conversation, either because what they built isn't working, or because what they were promised never showed up. 

Then you have adoption. A bolt-on can also leave adjusters with one more place to log in and no time back. When that happens, they abandon the tool, and you’re left paying for expensive software that no one uses.

Path two: Move slowly

Claims professionals are risk averse, and for good reason. If AI is working in the background, what does that do to bad faith exposure or file handling? Many have also been promised things before. They’ve moved from TPA to TPA, tried the tech-forward ones, and gotten the same results (with their name on the decision each time). With so many unfulfilled promises, it makes sense that the instinct is to watch how others fare with AI before making large investments. 

The case for it: Taking time is protection. If AI goes sideways, then moving with caution was the right call. Leaders who move slower also get the benefit of learning from others’ mistakes, so when it comes time to adopt the technology, they can do so more strategically.

The catch: It’s easy to fall behind. Competitors that adopt early can run claims at lower cost and with better results, which shows up in their margins as a competitive advantage. AI is also moving fast enough that every quarter of waiting makes the gap wider and harder to close. Starting later means catching up to a standard that is moving further and further ahead.

Path three: Partner with someone who’s already built it

The third path is to skip the build and outsource claims to a partner that already effectively runs claims with AI. Leaders who choose this direction have usually looked at what a build takes and decided they’d rather buy the result than run the project.

The case for it: You own the outcome without owning the build. Your team stays focused on results instead of development roadmaps, vendor management, and rolling out new tools to adjusters. You skip the engineering cost and the months of adoption work, because the technology and the people are already in place. 

The catch: Bringing on a partner introduces risk, as you forfeit full control. Handing your claims to a partner means putting your name on someone else's technology and judgment. If something goes sideways, you're the one on the hook. 

It’s often about more than technology. A claims leader told me recently that going with an AI-enabled TPA was clearly the right answer, but that he'd have to let several employees go, which is a difficult and heavy decision to make. Giving up control of the claims experience means trusting a partner with your results, your team, and your reputation.