Crawford & Company, a provider of claims management and outsourcing solutions, is putting new artificial intelligence tools through a formal review process before they ever touch a live claim. The Atlanta, Georgia-headquartered company adopted its claims innovation approach specifically so adjusters and claims specialists – not developers or executives – decide whether an AI tool earns a place in daily work.
David Wright (pictured), chief AI, data, and cloud officer at Crawford & Company, said the approach reflects how deliberately claims organizations need to move on automation. “If we’re trying to dig in on a particular modification to our claims process, we take representative folks from the areas we’re modifying,” Wright said. “We give them a prototype of a new interaction – different from how they’ve been doing it – and say: if AI were to help you on this interaction, is this good for you?”
Wright's team tests new AI-assisted workflows with cross-functional groups of adjusters before any wider rollout. “They could say no,” he said. “Or: is this accurate? That’s actually a really big one. We have to repetitively test and make sure that across multiple uses, it’s stable and accurate.”
Wright pointed to reserve-setting, a core judgment call on longer-running claims, to illustrate just how high the bar for validation would need to be. "Let's say that we wanted to find out if AI can help with reserve setting, which would be a really big deal," he said. "That's where this intensive validation approach would be particularly important. If it’s a longer-living claim, adjusters have to estimate what it’s going to cost and set those reserves. We'd want to be asking, ‘now how’s it doing? Is it doing well enough? You know what good looks like. Is it making mistakes? Are there concerns?’”
If adjusters flag problems, Wright’s team shuts the tool down and refines it until they approve it – a cycle that reflects how quickly artificial intelligence is reshaping the claims industry across the sector this year. “We’re doing that over and over again at scale, with many, many little workflow and process modifications,” Wright said.
Crawford’s caution is shaped by clients and regulators as much as internal preference, Wright said, amid the growing debate over how AI is reshaping claims department jobs carriers face as they pilot automation more aggressively. “The industry is very regulatorily dominated,” he said. “For a lot of claims adjusting – which is one of Crawford’s main services – a lot hinges on the license, the expertise, and the judgment of the adjuster.”
He said Crawford’s carrier and corporate clients are not asking AI to replace that judgment. “They’re not looking to replace the adjuster,” Wright said. “They want to augment the adjuster, help them be faster, and let them spend more time on the really hard parts of the claim.” He added that clients’ compliance cultures reinforce the position: “We’re paying for their judgment; they’re licensed and regulated where they work. They are still the ones in charge, the ones having to make the judgment.”
On the build-versus-buy question tied to how carriers are preparing their claims teams for AI, Wright said Crawford favors developing its own capability over third-party tools, even though it relies on outside large language models it cannot build itself. “We’re finding that our own primary engagement with tools, and the process of building our own skill and expertise, is the main emphasis,” he said. “We’re not going to outside tools as a preference – we’d rather engage with the changes ourselves so that we really understand what we’re doing, the logic of it, and where it’s leading.”
Wright said leading transformation means working from two directions at once – securing funding and buy-in from Crawford’s executive leadership while giving frontline workers room to experiment. Microsoft 365 Copilot is one tool employees can already use to build their own agents, he said. “If they say, ‘I created an agent and I like what this is doing,’ then I can come alongside them and say: let’s check the security, let’s check the privacy. Can we take your idea and promote it so other team members can benefit from it? It’s not just your idea anymore.”
That two-way approach, Wright said, is central to scaling AI responsibly across Crawford’s workforce. “What we really want is to enable hundreds and hundreds of workers to be the sources of ideas and innovation,” he said. “Getting tools in their hands and getting them into a mindset of experimenting and trying things is one of the most important things I work on, because over time that will multiply our innovation capacity.”
