Professional-grade AI for profitable underwriting.
A New York software company teaching commercial insurance to read its own paperwork — and decide what's worth covering.
EXHIBIT A: The underwriter's new desk. What used to be forty browser tabs now fits behind one login. Allegedly, the coffee still gets cold.
It is a Tuesday morning at a commercial insurance carrier somewhere. A submission lands: a restaurant in Ohio wants general liability coverage. A few years ago, this meant a half-day of detective work - pulling the loss runs apart, squinting at a website, guessing at the square footage, hoping the broker mentioned the deep fryer. Today the underwriter opens Kalepa's Copilot and the deep fryer is already on the screen. So is the prior claim nobody mentioned. So is a clean, structured read of a business that arrived as a messy PDF.
Kalepa is the company behind that screen. It builds AI software for the people who decide what insurers will cover and at what price - a job that has run on judgment, spreadsheets, and gut instinct for roughly two centuries. Kalepa's bet is that judgment is still the point. The forty tabs were never the point.
“The forty tabs were never the point. The judgment was.”
The Kalepa thesis, condensedCommercial insurance has a peculiar disease: the most valuable thing about a risk is usually the thing nobody wrote down. The application says “restaurant.” It does not say “restaurant with a rooftop bar, a history of slip-and-fall claims, and a lease that expired in March.” That information exists - scattered across loss runs, websites, public records, and the broker's vague email - but assembling it is slow, manual, and easy to skip when the queue is two hundred submissions deep.
So underwriters do the human thing. They triage on instinct, quote the easy ones, and let the genuinely interesting risks - the ones where money is made or lost - get the same fifteen minutes as everything else. The result is a portfolio shaped less by deliberate strategy than by whoever had time that week. Kalepa looked at this and saw not a staffing problem but a reading problem. The data was there. Nobody could read it fast enough.
“The most valuable thing about a risk is usually the thing nobody wrote down.”
Why Kalepa existsIt sounds like the opening of a joke. It is, in fact, the cap table. Paul Monasterio earned a PhD in computational physics and nuclear science from MIT - the discipline of modeling systems too complex and too dangerous to guess about. Daniel Hillman came from systems engineering at Penn and a stint leading intelligence work in the Israel Defense Forces - the discipline of turning scattered, unreliable signals into a decision someone has to stand behind.
Neither of them was an insurance lifer, which is either reckless or exactly the point, depending on how you feel about insiders. In 2018 they made a bet that looks obvious now and looked premature then: that the same machinery used to model reactors and parse intelligence could be aimed at the unglamorous, enormous business of pricing risk. They named the product Copilot - a deliberate word. Not autopilot. The underwriter keeps the controls.
MIT PhD in computational physics and nuclear science. Went from modeling nuclear reactions to modeling commercial risk, on the theory that both reward people who refuse to guess.
Systems science engineer (UPenn) and former intelligence lead in the IDF. Spent a career turning messy signals into decisions - which, conveniently, is the entire job of underwriting.
Plenty of startups promise to automate the human out of the loop. Kalepa promised the opposite. Copilot does the reading, the cross-referencing, and the scoring; the underwriter does the deciding - and can see why the machine flagged what it flagged. Transparency is the feature, not the footnote.
Marketing word choice as company philosophy, exhibit one.
Copilot is not a single clever trick. It is the entire underwriting workflow, modularized, with AI doing the parts humans are bad at - reading fast, reading everything, never getting bored on submission number 198. A messy email becomes structured data. A business gets classified to the right NAICS code. Loss runs get extracted. Hidden exposures get surfaced. A risk gets scored, and the score comes with its reasoning attached.
Classifies documents and extracts data from emails, applications, and loss runs - so nobody retypes a PDF again.
Detects conflicts, routes work, and prioritizes the submissions actually worth an underwriter's morning.
A unified view of exposures and controls, drawn from billions of structured and unstructured data points.
Plugs into pricing models and generates documents to move a qualified risk from “maybe” to bound.
Automated scores that show their work - the model explains the flag, so the underwriter can trust or override it.
Real-time optimization and guideline adherence across an entire book, not one policy at a time.
“Copilot does the reading. The underwriter does the deciding.”
How the product divides the laborPaul Monasterio and Daniel Hillman found Kalepa in New York and launch the first version of Copilot for commercial underwriters.
Inspired Capital leads the round, with IA Ventures and operators like Gokul Rajaram, Jackie Reses, and Henry Ward joining in.
Munich Re Specialty - North America, Bowhead Specialty, Canopius, Hinterland and others deploy Copilot across their books.
A 125-year-old mutual insurer picks Kalepa to bring AI into its underwriting - legacy meeting the new tooling.
Fourth consecutive year on FinTech Global's InsurTech100, plus a spot on the 2025 AIFinTech100.
Here is the test for any insurtech: do serious, conservative, deeply skeptical insurance companies actually run it? For Kalepa, the answer is a list. Munich Re Specialty - North America deployed Copilot for a deeper read on its exposures. Bowhead Specialty, fresh off its IPO, adopted it for risk selection. SECURA, a mutual insurer older than the airplane, chose it in 2025. Canopius and Hinterland signed on. More than fifteen carriers in total.
These are not companies that adopt software for the novelty. Bowhead's CEO Stephen Sills put it plainly: “No one has impressed us as much as Kalepa.” That is the kind of sentence an insurer says only after the procurement committee has finished trying to say no.
Illustrative split of effort, before vs. after a Copilot-style workflow
Directional illustration of Kalepa's value proposition, not audited metrics. The point: less hunting, more deciding.
“No one has impressed us as much as Kalepa.”
Stephen Sills, CEO, Bowhead SpecialtyInsurance is a promise: pay now, and we will be there when the worst happens. That promise only holds if the people making it can actually see the risk they are taking on. Mispriced risk does not just hurt a carrier's quarter - it erodes the thing the whole industry runs on. Kalepa's stated mission, “professional-grade AI for profitable underwriting,” is really an argument about trust: better-informed decisions make for promises that can be kept.
The company's culture follows from the work. It hires people who understand both deep technology and financial services - a rarer combination than it should be - and spreads them across more than ten countries. The discipline it sells is the discipline it claims to practice: don't guess, show your reasoning, keep the human accountable for the call.
Kalepa shares the field with insurtech names like Cytora, Akur8, Planck, Federato, Send, and Convr - plus the most stubborn competitor of all, the carrier's own in-house tooling. The differentiator Kalepa leans on is breadth: not one module, but the full submission-to-bind workflow, with the reasoning exposed at every step.
The hardest rival to beat is always “the way we already do it.”
Return to that underwriter and the Ohio restaurant. In the old version, the deep fryer stays hidden, the expired lease never surfaces, and the policy gets priced on a guess dressed up as expertise. In Kalepa's version, the risk arrives pre-read. The underwriter spends the saved hour on the thing only a human can do: deciding whether this is a risk worth taking, and what it is worth.
That is the change Kalepa is after - not fewer underwriters, but underwriters pointed at the part of the job that actually requires them. As more carriers wire AI into the point of decision, the quiet advantage goes to whoever reads the most, the fastest, with the clearest reasoning. Kalepa has spent since 2018 making the case that this is a reading problem with a solvable answer. The list of carriers running Copilot suggests the industry is starting to agree. The coffee, for the record, still gets cold.