Who Wins the $1TN Insurance Market?

Last week Forbes reported that Corgi, an AI insurance company founded in 2024, had raised again at a $4 billion valuation. Its fourth round of the year. In January it was worth an estimated $630 million.

Yesterday, a second one arrived. Ominimo, a Serbian-Hungarian startup founded in 2024 by three ex-McKinsey consultants, closed a round at a $1.6 billion valuation. Profitable, two years old, and being celebrated at home as Serbia's first unicorn.

Meanwhile Allianz is cutting up to 1,800 jobs at one of its subsidiaries, and its chief executive named the reason out loud: AI.

If you sell legal work, accounting, consulting or any other professional service, read this one as a preview. Insurance is further down the road you are on. It has AI-native firms taking real market share, incumbents cutting staff and blaming the technology, and private equity that spent a decade buying up the industry now discovering that AI reprices everything it bought.

This edition maps who wins. Same four positions as accounting... the tools, the AI-natives, the incumbents, the buyers.

Insurance numbers get thrown around loosely, so it is worth separating them properly, because the headline figure is not the one that matters.

You will see insurance described as a $7 trillion industry. That is premium volume, meaning everything policyholders pay in. Life insurance alone collected $3.1 trillion in 2024 and property and casualty another $2.4 trillion, according to Swiss Re, with health taking the total past seven.

But premiums are not revenue for work. Most of that money passes straight through the business... paid out as claims, or held back as reserves against future ones. Nobody is automating a claim payment away. It belongs to the policyholder.

What AI can take is the cost of running the business: the quoting, underwriting, broking, claims handling, policy admin and call centres that sit between the premium coming in and the claim going out.

Insurers spend about a quarter to a third of premiums on it in property and casualty, and rather less in life, where much of the premium is really savings. On P&C alone that is $600 to 700 billion a year. Add life and health administration and it comes to somewhere between one and one and a half trillion dollars. I'll call it a trillion and stay on the conservative side.

That trillion is the market this edition is about. On a par with legal services at $1.1 trillion, and comfortably bigger than accounting at $650 billion. And easier to automate than either, because insurance is built from the three things AI does best. Underwriting is prediction. Claims is document processing. Broking is matching.

The sell-software-to-insurers layer is crowded and well funded. Akur8 and Gradient AI price risk. Federato, Sixfold and Pibit rebuild underwriting. Reserv, Qantev and Wisedocs automate claims.

And insurers are buying. By one industry survey, three quarters of US insurers were already using generative AI in 2024. McKinsey puts the revenue opportunity at $50 to 70 billion.

But look where the money went. Over the past year, digital insurance brokers... companies that own the customer and the economics rather than selling software to whoever does... took 37.5% of insurtech deals and nearly 75% of the capital.

More deals going to the tools, but most of the money going to the companies that own the insurance.

My call: same trap I mapped in legal. Tools teach the machine how the industry works, and then the machine belongs to someone else. Selling software to insurers is a good business right up until the insurer builds it in-house, the platform bundles it for free, or the AI-native skips buying tools entirely because the tool is the company.

Two companies on two continents, built completely differently. Both doing the work rather than selling software to the people who do it.

Corgi. San Francisco, founded 2024, out of Y Combinator. It sells liability cover to startups, with AI generating quotes and processing claims instead of teams of human assessors.

Corgi writes through a Risk Retention Group, a pooled self-insurance vehicle. Lighter on regulation than a full carrier licence, heavier on the need for reserves.

A $108 million Series A in January, at a valuation PitchBook estimated at around $630 million. A $160 million Series B in early May at $1.3 billion. Three weeks later, another $106 million at $2.6 billion. And last week, Forbes reported a further round at $4 billion (Corgi has not confirmed yet).

The revenue explains some of it. Corgi said it was at $40 million annualised when the Series A closed. Sources told Forbes it is on track for $450 million by the end of the year.

The rest is just what insurance is. Insurance eats capital. Every dollar of growth needs reserves standing behind it, which is why an insurer that grows this fast raises this often.

Ominimo. A Serbian-Hungarian startup founded in 2024, selling motor insurance to European drivers.

Different structure, same idea: Ominimo is a managing general agent, so partner carriers like Signal Iduna and Zurich's DA Direkt hold the underwriting risk while Ominimo's models do the pricing and run the customer relationship. Part of the new money goes toward its own licence, which would let it keep the profit it currently shares.

The numbers are hard to argue with. Seven per cent of the Hungarian motor market and 300,000 policies in the first twelve months. Premium run-rate up twelvefold since 2024, now around $350 million. A valuation of roughly $230 million in May 2025, $1.6 billion now. And profitable, which almost nothing in insurtech has been.

The edge is old-fashioned: better pricing. Ominimo says it uses hundreds of data points where traditional motor insurers use five or six. The team is 130 people, two thirds of them in data science or engineering, including eight Maths Olympiad medallists.

Both companies sell finished insurance, not software. That is the whole distinction this newsletter exists to draw.

My call: the natives win where speed and pricing are the product... startups, motor, SME, the new risks incumbents will not touch. Their ceiling is capital and licences, both of which accumulate slowly. So watch two things: who rents them a balance sheet, and which of them gets a licence of its own first.

Start with the receipt. Allianz Partners, the travel and assistance arm, is cutting 1,500 to 1,800 jobs across Europe... up to 8% of the unit... through severance and early retirement. The unit handles around 200,000 calls a day, many of them questions a model answers instantly. Management did not hide the reason.

That is one subsidiary of one carrier. The structural problem is worse. Most carriers run core systems built in the 1990s, mainframes nobody wants to touch. Layering AI on that is archaeology, not product work, which is why, by one industry survey, 90% of insurance executives say the business needs reinventing and only a quarter have started.

So the big brokers chose the faster route: buy it. Gallagher paid $13.45 billion for AssuredPartners. Aon paid $13.4 billion for NFP. Brown & Brown paid $9.8 billion for RSC. Marsh took McGriff for $7.75 billion.

And the most telling one: in December, Willis Towers Watson paid $1.45 billion for Newfront, a tech-native broker. That is an incumbent paying over a billion dollars for something it could not build itself.

My call: the big carriers survive. They own the capital, the licences and a century of claims data, and they will shrink their way across... Allianz is showing how. The exposed seat is the mid-tier broker. Too big to run lean, too small to buy technology, and sitting on exactly the work models learn fastest: quoting, placing, renewing. The hole in the middle, again.

Insurance is different from every market I have mapped, because the roll-up machine is not arriving. It has been running for a decade, and it is the most developed in professional services.

Private equity sits behind roughly seven of every ten agency deals in America. The engine has run the same way for a decade... buy a local agency at 5 to 9 times earnings, fold it into a platform trading at 13 to 18 times, and the gap is value created by arithmetic alone.

Run that a few hundred times and you get Acrisure. From $38 million of revenue to about $5 billion in eleven years, roughly 900 acquisitions, a $32 billion valuation at its 2024 recapitalisation.

Now watch what it says about itself. The chief executive describes Acrisure as "an AI- and technology-powered global financial services provider." Its technology chief describes embedding engineers directly into live workflows... in his words, they build "in situ, from the experiential pain of real workflows."

That is forward deployed engineering... the AI labs' own playbook, running inside a brokerage. They have hired from Palantir. Business services is already 18% of revenue. They have also made AI-driven job cuts of their own.

The traffic runs both ways now. This month Mile Auto, an AI-native insurer, reportedly bought Insurance House, a 62-year-old agency, for its distribution. The incumbents are buying AI. The AI is buying incumbents.

Then there is the question PwC asks, which I believe is the right one: does AI let new entrants deliver brokerage at structurally lower cost, or does it let the consolidators cut costs across everything they already own? Their own read is that uncertainty about AI is already pulling brokerage valuations down and narrowing the arbitrage that powered the whole decade.

My call: the roll-ups do not stop. What they are buying changes. Yesterday an agency was a book of business and a commission stream. Tomorrow the same agency is client relationships, distribution and training data for an AI production line. The buyers who understand they are acquiring feedstock win. The ones still paying 2021 multiples for commission streams are buying melting ice.

The most interesting behaviour in this market is the insurers' own.

Research from AIUC, co-authored with researchers from Anthropic and OpenAI, found that more than 90% of insurers' exposure to AI sits in "silent" cover... buried inside ordinary policies, unpriced, often unnoticed.

Instead of pricing it, they are backing away from it. New exclusions for AI risk are moving through the industry, and lawyers in the field say carriers are moving fast to limit the exposure.

There is a small counter-market forming at the same time. AIUC and Mount underwrite AI agents. Lloyd's launched cover for losses caused by hallucinating models. These are insurers selling policies on AI, not AI companies selling insurance. Different quadrant, opposite direction.

And the demand is arriving. Crosby, an AI-native law firm, said this week it is working with auditors, regulators and insurers to get professional liability cover for its agents, so they can do legal work without a lawyer reviewing every output. Today its lawyers check everything. By their own account, the thing stopping them is not capability. It's liability.

Which puts the underwriter, not the lab, in charge of deciding when the human comes out of the loop.

In the accounting edition I said the AI-powered roll-ups would turn toward insurance brokerage within a year. That still looks right to me, and the capital is already moving that way.

I would expect a frontier lab, or a vehicle it holds equity in, to take a stake in an insurance distribution or claims business inside the next year. That is the accounting pattern, and insurance is the obvious next stop.

Within about eighteen months, I think one of the big consolidators buys an AI-native carrier or MGA rather than competing with one. Cheaper than building it, and they already know how to buy things.

And within a couple of years, AI liability should be its own line of business at the major carriers instead of an endorsement tucked inside cyber cover.

Legal showed who's fighting for the work. Accounting showed the buyer. Edition three showed the labs coming down to do it themselves. Insurance shows what happens when they arrive in a market private equity already owns.

If you work in insurance and I have missed something you see from the inside, reply... I read everything.

See you next Tuesday.