A hurricane makes landfall and the claims operation gets its hardest test of the year. This time it gets it with fewer people. Sedgwick expects about a quarter of adjusters to leave the business by the end of 2027, and Verisk has catastrophe claims at 43% of total assignments, up from 34%. Fewer claims overall, but more of them arriving at once.
I sent questions to Tom Freeland, President of Liberate, as Hurricane Isaias was heading for the coast. Tom has spent more than 20 years in P&C, much of it in Florida and Gulf markets, including running a parametric hurricane insurer. Liberate's carrier customers saw a 14x jump in call volume over two weeks during Helene and Milton. (Earlier this week we ran Liberate CEO Amrish Singh on AI agents calling insurers.) His answers are below, lightly edited.
When a hurricane makes landfall, what's the first thing that breaks in a carrier's claims intake? Phones, FNOL forms, triage, adjuster assignment?
Voice phone reporting breaks first. Before an insured calls, they have lived through a pressure cooker: gathering documents, filling sandbags, hunting for gas, driving hours to a hotel. At Florida Blue and GuideWell I learned that a hurricane is the largest mental health event of the year. Then they come home to no power, no internet, and a house they can't live in. This isn't a fender bender in a parking lot.
They call wanting to tell their story and get a claim number and next steps. Instead it can take several calls over one to three days just to reach a person, and then the rep is trying to get the basics and move on. Regional carriers' own employees are hit by the same storm, and offshore CAT providers usually aren't integrated, so they can't start anything for the insured. Reports are thin, the bottleneck moves to the adjuster, and cycle time suffers. Speed drives both satisfaction and what the carrier pays. Heroics don't scale either. I know a CEO who claims calls himself. He got so absorbed that calls ran long and he paid claims on the spot.
An AI voice agent changes that. Nicole, our persona, never puts anyone on hold, is never rushed, and captures a complete report. With deep carrier integrations she can also dispatch water mitigation, so blue tarps go on roofs faster. That is empathy, and it limits damage.
Your carrier customers saw a 14x jump in call volume over two weeks during Helene and Milton. What did those calls actually look like, and how many were true FNOLs vs. status checks and "is my claim open" calls?
Not every call in a surge is a policyholder reporting a loss. Many are lenders verifying coverage, lawyers reporting for claimants, and AI systems placing calls of their own. Among policyholders, many call because flood isn't covered by their homeowners policy, or because their hurricane deductible is a percentage of the home's value.
Sedgwick expects about a quarter of adjusters to leave by the end of 2027. If carriers can't hire their way through a surge anymore, what's the realistic playbook?
I won't predict how far AI goes in adjudication, but I can speak to the work around it. Imagine an auto claim where AI starts downstream work at first report. It lines up the wrecker, ride and rental, routes the car to a well-rated body shop that isn't backlogged, finds an orthopedist if the insured needs an MRI, pulls the police report, witness statements and vehicle value, and sends it all to the carrier's fraud model. The adjuster then does what only they can: adjudicate and own the negotiated outcome. They can serve far more people, and the job gets better.
The talent problem starts at FNOL, a revolving door with average tenure around nine months. The best get promoted to adjuster, and the rest burn out of a job that is boring until it's overwhelming. Few executives ever listen to FNOL calls. A $5 FNOL looks cheap until you count recruiting, training, turnover, and the cost of getting it wrong.
Verisk has CAT claims at 43% of assignments, up from 34%. Fewer claims overall but more arriving at once. How should a claims leader plan staffing around that?
You can't afford to staff for peak and still keep your expense ratio low enough to price competitively. In auto especially, price wins, and AI will make rate shopping even easier.
Events are also getting faster and harder to predict: tornadoes, hail, lightning, fire with wind. The most impactful one I lived through was the April 27, 2011 outbreak in Alabama, with 62 tornadoes in about 18 hours and more than 230 deaths. I was speaking in Florida while my wife and kids were home in Birmingham. Carriers weren't staffed for it, and the same was true in the California and Hawaii fires and the recent Midwest hail. The answer is capacity that scales instantly. AI that can take thousands of calls in the same second is what's needed, and not all AI vendors are equal.
Where should AI own the intake work outright, and where does it need to hand off to a person? Give us a concrete example of each.
Let AI own the calls that don't need human judgment, with us involved in routing. Examples: a lender verifying coverage, a lawyer reporting a claim for a claimant, and AI systems calling in claims. AI-to-AI calls can bog down a call center, so they should go to AI. Auto glass is another. AI can send the customer to an approved provider and book the appointment, just as a consumer would walk into the shop and pay their deductible. The same goes for dispatching water mitigation the moment a roof leak is reported.
Hand off to a person for injury or death, even at FNOL. Our 70+ carriers choose how each claim type is handled, and we recommend an immediate transfer for serious injury. If the carrier isn't staffed at that hour, we follow a prescribed path to collect information that speeds the route to human help.
What's a mistake you've seen carriers make with automated intake during a CAT that made things worse for policyholders?
Using AI that isn't grounded in insurance. People who know insurance and human conversation need to be involved from the start, and a vendor that does AI across many industries isn't the one to get this right. Guardrails are mission critical, with a systematic way to catch hallucinations. What happens when the AI takes a report and never sets it up in the system? Conversation is the easy part. Many vendors know technology but not insurance, and others come from insurance but have no active deployments.
The other mistake is build versus buy. I've watched teams build, spend heavily, and return a year later asking us to deliver. We can get many use cases live in 8 to 12 weeks. AI decisions are also increasingly made by technologists rather than the business leaders who own the P&L and answer to the Board. The point of AI is competitive advantage, and the person who chose to build won't be the one explaining lost market share. My advice: build your own models for pricing and fraud, and buy a tested product for intake. It's also far cheaper.
You ran a parametric hurricane insurer. What did that teach you about claims that traditional carriers could borrow?
Parametric was a great idea and truly transformative, but adoption was hard. Personally and financially, it was my biggest failure. The product gave a declared amount of coverage on a property. If a storm of a set strength passed within a set distance, payment went out within 24 hours. A $25,000 policy with a Category 5 eye within 25 miles meant $25,000 quickly on its way. A Category 2 at 100 miles might pay $5,000. It covered what homeowners policies don't: docks, pool screens, landscaping, lost wages.
Carriers can borrow the concept to cover deductibles and uncovered items, which would make claims faster and far less frustrating. The hard part was getting people to pay for it when home insurance is already so expensive. The more recent a bad hurricane year, the easier the sell.
For carriers in a hurricane's path, what's the one thing they should be doing in the 48 hours before landfall?
Proactive communication. Call or text your customers before they call you. Our tool can hold an AI voice conversation once the storm passes to collect information, which changes how reporting works.
Support two groups. Your agents sold the policies, may be affected themselves, and often have personal relationships with customers. Your newer adjusters will spend days explaining that flood isn't covered, or that a hurricane deductible, commonly 1 to 10 percent of the home's value, isn't $250. The crying on the other end never becomes routine, so give them a support network.
Tom Freeland is President of Liberate. Connect with him on LinkedIn.