Greg Cornett has been in this business long enough to remember when the office ran on Windows 95. As AVP of Claims at Harbor Claims, a Florida-based operation handling claims across the US, he is exactly the kind of person a new technology has to win over: experienced, accountable for getting coverage decisions right, and not easily impressed.
So when Harbor first sat down with an AI vendor, Greg didn’t come to be sold. He came to break it.
“I purposely went into our first-ever meeting with an AI vendor looking to break it… it very much exceeded my expectations. I was a little blown away.”
That arc, skeptic to believer to in-house champion, is the most useful thing about Greg’s story, because he can tell you exactly what made the difference. Speaking on the “AI Reality Check” panel at Insurtech Insights 2026, alongside Voltaire, the tool Harbor now runs, he laid out something close to a playbook. Here it is, in his words.
1. Try to break it first
The fastest way to trust an AI tool is to attack its hardest case, not its easiest. Greg was handed a softball: write a payment letter. He lobbed a grenade back.
“I was given a ‘let’s write a payment letter.’ Instead I threw out, ‘let’s write a partial denial letter.’ And it did it. It pulled the policy language I wasn’t expecting it to pull. The results were, frankly, a little astounding.”
A partial denial is one of the hardest letters in claims. It has to be precise about what’s covered, what isn’t, and exactly why. And it has to hold up if someone challenges it later. Watching the tool handle the edge case instead of the demo case is what flipped him. If you’re evaluating AI for claims, don’t grade it on the easy letter. Hand it the one your best adjuster sweats over.
2. It starts at the top
Ask Greg for the single biggest predictor of whether AI sticks, and he doesn’t reach for technology.
“Top-down support is critical. It’s absolutely the most important thing. You have to have ownership and leadership that believe in this process… if we’re not on board with this, we’re going to get left behind.”
Harbor’s advantage was a forward-thinking ownership group that pushed from the very beginning, which meant the question was never whether to adopt, only how well. Where that conviction is missing, pilots tend to die quietly in the budget cycle. It’s the same top-down conviction Heritage Insurance’s director of claims credits for making AI stick. AI adoption is a leadership posture before it’s a tool.
3. Sell it as support, not replacement
The fear in any claims shop is that the tool is there to replace people. Greg watched that fear turn into something else entirely.
“They understand this is not here to replace them. This is here to support them. And if it makes their job easier, they’re all for it.”
The tell that adoption has taken hold isn’t a usage dashboard. It’s the change in the questions people ask.
“Instead of ‘how can I avoid using this,’ it becomes ‘could you maybe add this letter too? Can you do this for me as well?’ That light-bulb moment is always fun to see.”
That shift didn’t come from a mandate. It came from adjusters discovering the tool handled the tedious part and left the judgment to them, which is exactly the point. The AI sits next to the adjuster, not in their chair.
4. Invest in your people in parallel
AI didn’t replace Harbor’s talent strategy; it ran alongside a bigger bet on people. With the industry short tens of thousands of adjusters, Harbor’s answer was to build its own bench through Claims College.
“Part of the problem with onboarding new adjusters is the overwhelming amount of information thrown on them… So we started Claims College… Does it take longer to get them on live claims? Yes. But once we do, they’re more efficient from the get-go, a lot less standing up over the cubicle, ‘hey, how do I do this?'”
The tool and the training compound each other. New adjusters lean on consistent, policy-grounded drafts while they build their own judgment. The payoff: fewer over-the-cubicle questions, a faster ramp, and a higher floor on quality from day one.
5. Start with the language-heavy grunt work
Asked where a carrier should actually begin, Greg points at the work that is almost entirely reading and writing and almost no judgment.
“QA review is low-hanging fruit… It’s so simple to go in and just present the claim file and say, ‘could we have done this better?’ And it’s amazing some of the results we’ll get.”
Claims correspondence and QA are language tasks at heart: high volume, high consistency demands, low tolerance for error. That’s where the tool earns its keep first. And once it’s in the file, it starts surfacing things no one had the hours to catch:
“AI might find a trend that a human might not recognize… put a stop to bad habits or implement corrections mid-flow instead of waiting for that three-month delayed review.”
6. Aim higher than speed: get it right
This is where Greg’s story stops sounding like a productivity pitch. The most valuable thing the tool does at Harbor isn’t writing faster. It’s occasionally stopping a decision.
“We have a saying in our company: if you’re trying to find the policy language to support a denial and you can’t find it, there’s a pretty good chance you shouldn’t be denying that.”
That principle is built into how the drafting works. As Voltaire CEO Yo Sub Kwon put it on the same panel: “If someone is trying to deny some aspect of a claim and there’s no relevant policy language that supports that denial, the software won’t let them, it kicks it back.” When there is supporting language buried in the policy, the draft pulls it forward, and adjusters have said they’d never have thought to cite it. The result, in Greg’s words, is a tool that helps a good adjuster “reconsider their position” on a claim before it goes out the door.
Speed is the obvious win. The quieter one is a tool that helps a good adjuster catch the denial they shouldn’t make, which, in claims, is worth more than any cycle-time number.
The Monday-morning takeaway
Greg’s one-sentence advice to the room was characteristically blunt:
“AI is a short-term investment pain point, but a long-term benefit.”
He’d know. He came to break it. He stayed to build his operation around it.
See how Voltaire helps adjusters draft claims correspondence
Harbor Claims runs Voltaire to help its adjusters write faster, more consistent, policy-grounded claims letters, with the supporting language surfaced and the gaps flagged before a letter goes out. Request a demo to see what it looks like next to your team.
