Many insurers are issuing AI press releases. It’s all very exciting. So, as a service to the investing public, at Lemonade we built an AI BS detector: three KPIs, public data, no spin. It won’t boost our popularity. It might boost your returns. Get the whole story at https://lnkd.in/eg4ReTDT
Fine and fair article. However, since you were founded 10-11 years ago, the top 10 personal auto insurers in the USA are still the top 10 in just about the same market share order. Same in homeowners. While your combined ratio is improving from 197% last year to 167.7% this year you have a long way to go to see a profit. Perhaps you will get there one day. Thankfully though, property and casualty insurance is not books, refrigerators, vitamins and tools. Discount the incumbents all you want, they are not going away anytime soon or ever.
It’s easy to buy business. Their loss ratio is trbl and they don’t have a secret UWing sauce.
"For these reasons market leadership has shifted every time a substrate changed. In each case, the technology was visible and accessible. Incumbents invested. Analysts cheered. Press releases said all the right things. And yet leadership reordered. It’s not that incumbents were oblivious. It’s that they were maladapted." Except, this conveniently ignores that, in the insurance business, the best adapter was not a startup, but rather Progressive. The incumbent ate all the other incumbents and the new challengers. The second best adapter was GEICO, another incumbent. The only semi-successful challenger model was Esurance which never made a profit and got swallowed by Allstate who did very little with it so it basically went extinct. Cool story, though!
Agree. When the downstream data is not reliable how can AI be applied? When the products downstream aren’t fully digitally enabled how can they be ready for MCP apps? Foundation is needed first to scale.
What’s the combined ratio of lemonade? 180-200%. Combined ratio is like golf, low score is better. Maybe lemonade will understand that some day, or maybe not. If this was basketball they’d be killing it, not golf.
The interesting shift with AI in insurance isn’t only about operational efficiency or growth metrics. As underwriting and claims decisions become increasingly automated, the harder question for the industry may become: can the exact decision conditions be demonstrated when a claim is paid, denied, or reserved? When disputes arise months or years later, investigations often focus less on the model itself and more on what information, signals, and authority existed when the decision occurred. As AI adoption expands across the sector, the ability to evidence that decision context may become an important part of institutional trust.
Daniel Schreiber — architectural advantage and incentive alignment absolutely matter. But insurance is not primarily a scaling contest. It is a regulated capital allocation business. In P&C, superiority is tested less in growth curves and more in how underwriting and claims decisions withstand audit, litigation, and adverse development over time. AI-native architecture can accelerate operations. Execution accountability determines institutional durability. Scaling without proportional labor is a signal. Governance at bind, pay, and reserve movement is the constraint. The companies that win long term will have both.
Because it's digitalizing and complicating a process to ensure lack of paid claims to ensure premiums are justly pure profit with zero "insure" just a supplemental policy that will pay 25% value of your claim is that their benchmark internally is plainly not insurance. Imagine a company harrasing you to replace a roof when it was never inspected or seen yet put me in a corner to spend 25k because I could not find the record. I heard LEMONADE is the main insurance co in Pacific Palisades disaster.
An AI BS detector for insurance is long overdue. The gap between what carriers say AI is doing and what it's actually doing on the claims side is massive. From the restoration and claims defense perspective, the real question isn't whether a carrier has AI — it's whether that AI is being used to process claims faster and fairer, or to automate denials and scope reductions at scale. Three KPIs and public data with no spin is exactly how the industry should be evaluating these tools. The contractors and PAs who understand how carrier AI actually works — not the press release version — will be the ones best positioned to build files that survive automated review.
“Three KPIs, public data, no spin.” Doubtful. But selecting which public data counts is already spin. A 6% LAE on small renters claims is not comparable to bodily injury or commercial liability. A +320% improvement in profit per exposure is less revealing when the starting point is close to zero. And calling cash-burn-driven layoffs “AI-powered scaling” may be clever messaging, but it is not serious analysis. The real story is in the omission: nine years after launch, not one profitable quarter. Over a billion dollars of cumulative losses since IPO. Meanwhile, the legacy insurers dismissed here produce more profit in one quarter than Lemonade has generated in its full history. Any “AI BS detector” that leaves out profitability is not a detector. It is a prop. Public data. No spin.