Software can now carry an insurance risk to a bind-ready quote. A Nasdaq-listed carrier is the test case, and the combined ratio is where to look.
getty
Insurance was born in a coffee house. In the 1680s, merchants and ship-owners gathered at Edward Lloyd’s place near the Thames to wager on which voyages would come home and which would not, and the modern industry grew out of those bets. Three and a half centuries on, the trade still rests on a single human act which is that an underwriter looks at a risk and decides whether to take it, and on what terms. A New York firm called Sixfold now says a machine can do most of that work, right up to the quote a broker can sign.
On June 15, Sixfold opened its AI Underwriter to property and casualty insurers, a launch first reported by the trade title The Insurer from a company statement. The pitch is straightforward enough. The software learns one carrier’s book and its appetite for risk, reads each new submission as it arrives, and recommends what to do next. Set up for what the trade calls straight-through processing, it can take a case all the way to a quote-ready and bind-ready state, the point at which a broker or client can simply say yes. A human underwriter still signs off and still carries the can. The novelty is how far the machine now travels before the human is needed.
Why “Bind” Is The Word That Counts
For years, AI in insurance meant a clever assistant. It read documents, flagged the odd anomaly, and saved an underwriter an hour of slog. To bind a policy is a different order of thing. Binding is the moment an insurer commits its own capital to a risk, the act that turns a quote into a promise to pay. Software that can walk a submission to that threshold is no longer fetching the coffee. It is standing at the table.
That matters for the people who own insurers, because underwriting is where an insurance company actually makes or loses money. The trade measures it with the combined ratio, the share of premium eaten up by claims and costs. Under 100% and the underwriting turns a profit; over 100% and it bleeds. Anything that lets a carrier write more business, price it better and waste fewer underwriter hours feeds straight into that number, and the combined ratio is the figure shareholders watch above almost any other.
The Listed Guinea Pig
Most claims about insurance software die in a fog of private companies and unaudited boasts. This one has a public test case. Among the carriers Sixfold names as customers is Skyward Specialty, which trades on the Nasdaq under the ticker SKWD, so its numbers are filed with regulators rather than printed on a sales deck. In its first-quarter results for 2026, lodged with the Securities and Exchange Commission in May, Skyward reported $667.7 million of gross written premiums, up nearly 10% on the year, and a combined ratio of 89.5%. That is a genuinely profitable underwriting business, the kind of disciplined shop whose endorsement of a tool is worth more than any vendor testimonial.
Sixfold says its software handled much of the underwriting analysis on submissions in Skyward’s excess-and-surplus property business during testing before the launch. John Burkhart, who runs property and casualty at Skyward, said in the announcement that the carrier was folding the agent into its underwriting while keeping, in his words, human expertise firmly at the centre. Read that as the polite version of a real boundary: the machine does the reading, the human keeps the pen.
Mind The Vendor Numbers
Now the caveats, and they are not small. Almost every eye-catching figure attached to this launch comes from Sixfold itself, measured by Sixfold, and audited by nobody. The company says customers have seen processing times fall by anywhere from 50% to 97%, hit ratios climb by 15% or more, and gross written premium per underwriter rise by up to 30%. A range that runs from 50 to 97 is not really a statistic; it is a span wide enough to drive a lorry through, and it usually means the flattering deployments are doing the talking. Treat the lot as the company’s own marketing until an independent customer puts its name to a number.
One claim is checkable, and it is the one that matters for scale. Sixfold says its customers represent about $270 billion in combined gross written premium and include Zurich, Generali’s corporate arm, Guardian, Axis and New York Life alongside Skyward. Those are serious houses, and a serious roster is itself a kind of evidence, whatever you make of the percentages. The firm was founded in 2023 and has raised north of $50 million, including a $30 million round in January led by Brewer Lane with backing from the insurance-software group Guidewire.
The Reason Carriers Are Actually Buying
Strip away the hype and there is a hard demographic reason insurers are reaching for this kind of tool. Underwriting is greying. Roughly one in four insurance workers in the United States is 55 or older, and the trade press has spent two years passing around a projection that the sector could lose about 400,000 people to retirement by the end of 2026. When a 25-year veteran walks out, decades of judgment about which risks to touch and which to leave go with them, and that knowledge is the hardest thing in the business to write down. Carriers cannot recruit their way out fast enough, so they are trying to bottle what the veterans know before it leaves the building.
That is the honest case for an agent like Sixfold’s. It can capture how a particular underwriter weighs a risk and apply it across thousands of submissions a junior hire could not get through. The catch sits in the same place. Younger underwriters often lack the instinct to spot when a model has produced a confident piece of nonsense, and an agent that quotes and binds removes some of the friction that used to give a wary human time to object. Speed cuts both ways, and the firms selling speed rarely dwell on the second edge.
Who Carries The Can
The harder question is not whether the software works but who answers for it when it errs. An underwriting agent that can quote and bind is making, or all but making, a capital decision. Bind a bad risk and a real insurer pays a real claim. Sixfold’s answer, sensible as far as it goes, is that each carrier decides how much the agent does on its own, and that the underwriter stays accountable for how it is governed. The company also says each insurer’s book is walled off, so one carrier’s data never trains a rival’s, which matters in a business where appetite for risk is a trade secret.
Regulators have not yet said much, though they will. American insurance is policed state by state rather than from a single federal desk, and supervisors who already scrutinise pricing models for hidden bias are unlikely to leave an agent that quotes and binds to its own devices for long. The National Association of Insurance Commissioners has spent the past two years circulating guidance on how carriers should govern AI in decisions that touch policyholders. An agent that can carry a risk to the brink of a binding quote is exactly the sort of thing that guidance was written to catch up with.
The wider current is unmistakable. AIG built a gen-AI underwriting assistant with Anthropic and Palantir and has pushed it across several lines. Brokers and carriers are wiring AI into the dull middle of the business, the reading and sorting and pricing of risk, faster than the rules can keep pace. Insurance spent two decades digitising its filing cabinet without changing much underneath. The cabinet is now starting to make decisions.
For an investor weighing a name like Skyward, the takeaway is concrete. Watch the combined ratio and the premium written per underwriter over the next few reporting periods, and watch whether they move in the right direction without the loss ratio creeping up to pay for it. If the AI is doing what its maker claims, that is where it will show up, in audited figures rather than press releases. Edward Lloyd’s punters at least knew they were gambling. The question now is how comfortable insurers, and the shareholders who own them, are letting a machine help place the bet.
