On October 19, a company called Mandl launched with a promise to rebuild the full engine of commercial insurance around AI. Mandl is a parent company, not a single insurer: its launch announcement names three operating businesses — Nirvana, an insurer its founder helped start; Vega, a claims administrator; and Cooper, an operations platform sold to agencies, brokerages, MGAs, and carriers. The founder is Rushil Goel, who cofounded Nirvana after leading Samsara’s fleet platform, and the release says Mandl is backed by General Catalyst, Lightspeed Venture Partners, and Valor Equity Partners.
The number your clients will repeat back to you is four. In the launch release, Goel says Vega is “already delivering cycle times four times faster than the industry standard.” The release does not define cycle time, name the industry standard, give a baseline in days, or point to anyone outside the company who has measured it. As published, four times faster is a claim by the company selling the service — not a finding, and not an audit.
Treat the claim fairly, though, because if it holds, it matters to a client’s cash. A commercial claim is working capital frozen in someone else’s process: the truck in the shop, the roof half-repaired, the deductible already advanced. Every week a claim sits open is a week the business carries a loss it already paid premiums to transfer. A claims operation that genuinely resolved claims in a quarter of the usual time would hand that float back — and business owners, who buy insurance for exactly that moment, will ask their broker about any headline that promises it.
Here is what faster does not change. Whether a loss is covered is decided by the policy — the grants, the exclusions, the conditions, the endorsements your client signed. A model processing that language faster produces a faster answer, and the answer can be no. Cycle time measures how quickly a claim moves; it says nothing about how broadly the form responds. A broker who sells speed as if it were coverage is writing next year’s complaint letter in advance.
Also say plainly what kind of evidence this is. Mandl the parent company is a day old, and the four-times figure arrives with no published methodology behind it. The operating companies are not a day old — Nirvana has written commercial trucking insurance for years, and the release puts it above $300 million in run-rate premium, a figure that is also the company’s own. None of that makes the speed claim false. It makes it ungraded — and a broker who repeats an ungraded vendor number to a client owns that number on the day a real claim tests it.
So the broker’s job is five questions, asked before business is placed, not after a claim sours. First: whose paper stands behind the policy, and who is accountable when the model gets a claim wrong? An AI workflow can recommend, but a carrier with a name, a rating, and capital has to own the decision. When Nirvana launched in 2022, trade coverage described its policies as written on state-admitted, A.M. Best A-rated paper backed by a panel of reinsurers — get the current version of that answer, for any AI-native market, in writing.
Second: what does the human review path look like when your client disputes an outcome? Every claims process denies and undervalues sometimes; the professional question is how a denial gets appealed, who reviews it, and whether that reviewer can actually overrule the model. Third: how are errors and denials audited — sampled by humans, overturn rates tracked, results shared with the brokers placing the business? A vendor that cannot describe its own audit trail is telling you the broker becomes the audit trail.
Fourth: what happens to the client’s data? Models like these run on operational data — telematics, schedules, loss history — handed over to earn a price. Ask who owns it, whether it trains models used for other customers, and what comes back if the client leaves. Fifth: what does month thirteen look like? Launch pricing and launch attention are not renewal pricing and renewal service, and who picks up the phone after the first renewal predicts the relationship better than any launch metric.
None of this is new in kind. Every insurtech wave has arrived leading with speed — quotes in minutes, claims in days — and some of those claims proved real while others quietly repriced at renewal or thinned out at the first hard loss. The underwriting takes years to grade; the marketing takes a day. Brokers have been the translation layer in every wave: the person who tells a client which numbers are measured, which are only promised, and which questions to ask before the difference matters.
So what do you tell the client who forwards you the four-times-faster headline? Tell them it’s a serious group with real backing and an operating insurer behind it, that the speed figure is the company’s own and publicly unproven, and that speed was never the part of the contract that pays a claim. Then tell them you’ll ask the five questions and report back. That conversation — not the headline — is what a broker is for.