Insurtech promised to disrupt insurance as fintech had shaken up banking. But one does not disrupt the bearing of risk as one disrupts an interface. The promise was neither kept nor betrayed, it was transformed, from conquest toward infrastructure.
In the mid-2010s, a promise gripped the world of insurance, that technology would disrupt the sector as it had upended banking, commerce and transport. Under the term insurtech gathered a generation of companies convinced that insurance was at bottom merely an old, badly written piece of software, which one only had to rewrite with smooth interfaces, algorithms and direct-to-consumer sales. The argument was seductive. Insurance rests on statistics and probability, terrain a priori ideal for automation, and its traditional players suffered from an image of heaviness, slowness and mistrust. The founders of insurtech promised insurance that was instant, transparent, friendly, free of the broker and the paperwork, and carried by a brand rather than by a network of agents.
The founders' mental model came largely from fintech, which had indeed shaken up retail banking. But the analogy carried within it a flaw of reasoning. Retail banking is for a large part an information and distribution activity, where technology can disintermediate the branch, accelerate payment and lower costs without transforming the nature of the product. Insurance, by contrast, is not only distribution, it is above all a promise to pay a future and uncertain loss, backed by capital and by risk pooling. To disrupt the distribution of insurance was within the reach of software, to disrupt the bearing of risk was not. By importing the fintech template without this distinction, insurtech confused the thin and visible layer of the trade, the interface, with its thick and invisible one, the balance sheet.
The markets embraced this promise with fervour. Lemonade, flagship of direct sales, goes public in July 2020 and sees its share, introduced at 29 dollars, exceed 180 dollars in early 2021, bringing its valuation to around 9 billion dollars while it held only around 200 million dollars of in-force premium1. Root, an auto insurer founded on driving data, lists the same year at a valuation of nearly 6.75 billion, higher than the market capitalization of the reinsurer Scor, established and profitable though it was2. Hippo, a specialist in connected homes, targets a valuation of several billion. Global insurtech funding peaks in 2021, with around 17 billion dollars invested over the year and a quarterly high close to 5.3 billion in the final quarter3. The promise then seemed not only credible, but imminent.
The reality of the insurer's trade reminded the disruptors of itself through an indicator that no interface can embellish, the combined ratio. This ratio relates the sum of claims and expenses to the premiums collected, and any level above 100 means the insurer is losing money on its underwriting, however elegant its app. Yet the pioneers of insurtech displayed for years combined ratios well above this threshold, revealing that they were certainly acquiring customers brilliantly, but insuring them at a loss. Technology had solved the problem of distribution without solving that of underwriting, which is the real heart of the trade. One could win over the policyholder in a few clicks and nonetheless be wrong about the price of the risk they were transferring.
This difficulty was not cyclical, it stemmed from a structural advantage of the established players, their data moat. To price a risk supposes a loss history long and deep enough to estimate the frequency and severity of losses, and the large insurers hold decades of data where the newcomers started from a blank page. To grow fast, the insurtechs underwrote portfolios they knew poorly, and the law of large numbers, which protects the patient insurer, turns against the hasty one. The trade they claimed to reinvent rested precisely on the resource they did not possess, the accumulated experience of risk8. The promise of disruption thus ran into a wall that software alone could not cross.
To this data moat was added an economic trap specific to the direct-sales model. By forgoing agents, the insurtechs had to buy their visibility through digital advertising, whose cost of acquiring a customer rose sharply as competition intensified. Yet personal-lines insurance is marked by fragile loyalty and high churn, so that a customer expensive to acquire may leave before being made profitable. The model assumed that the lifetime value of a customer would exceed their acquisition cost, an equation that did not hold as long as loss experience remained poorly controlled. Root experienced this most brutally in auto insurance, a hypercompetitive segment where its ratios long forbade any durable profitability, forcing it into a retreat and a relapse into loss after a brief profitable year7.
The euphoria broke sharply. The rise in interest rates and the general correction of technology valuations ended, as early as 2022, investors' tolerance for prolonged losses in the name of future growth. The revenue multiples of listed insurtechs collapsed, and private funding dried up in their wake. Global quarterly funding fell back into a range of one to one and a half billion dollars from early 2023, far from the 5.3 billion peak of late 2021, and the third quarter of 2025 totaled only around one billion, down 17% year on year3. Hippo's market capitalization contracted by more than 95% from its listing valuation, and the giant funding rounds, which exceeded thirty in 2021, fell to a handful4.
Beyond the figures, it was a narrative that was collapsing. The story of an inevitable disruption, which saw young start-ups replacing century-old insurers, gave way to a soberer observation, that of companies struggling to reach profitability in a trade they had underestimated. The number of insurtech unicorns, which had peaked around fifty-eight worldwide, shrank as several players went public at reduced valuations, were acquired or lost their status in down rounds4. The promise was not merely delayed, it was requalified. Insurtech discovered that it had not reinvented insurance, but that it had run into its deep nature.
The public market acted as an implacable judge. Several insurtechs had chosen the fast route of listing through merger with a special-purpose acquisition company, those vehicles that allowed a market debut without the filter of a classical offering. When enthusiasm subsided, these structures left shareholders heavily in deficit, and the mistrust spread to the whole sector, closing the window of new listings for a time. The market sanction preceded and amplified the private-funding sanction, because venture investors calibrate their valuations on the multiples of comparable listed companies. The fall of the latter therefore mechanically depreciated the former, propagating the public-market correction toward the private market and accelerating the selection between viable models and the rest.
From this correction was born a silent but decisive mutation of the model. The shrewdest survivors stopped wanting to replace insurers and set about serving them. The center of gravity of insurtech shifted from disruption to infrastructure, that is, from direct-to-consumer sales to the supply of technological building blocks to established players, and toward the underwriting-agency models, the managing general agents, which place risk without bearing it on their balance sheet9. The acquisition by Munich Re of NEXT Insurance for 2.6 billion dollars, a valuation in clear retreat from past ambitions, and the listing of Accelerant, a platform connecting specialist underwriting agencies with investors, for 3.4 billion dollars, illustrate this shift toward an insurtech of infrastructure rather than of conquest.
The appeal of the underwriting-agency model lies in its very economics, which inverts the difficulty of the direct-sales model. The agency places the risk, collects a commission and leaves the bearing of capital to a third-party risk carrier, generally a reinsurer or an established insurer. It thus captures the technological and distributive part of the value chain, the one where its tools excel, without immobilizing capital or bearing the volatility of loss experience, the one where the pioneers had failed. This decoupling between intelligent distribution and the bearing of risk allows growth that is less capital-hungry and profitability that is faster, provided underwriting quality is there. It enshrines a division of labor in which insurtech supplies the brain and the insurer the balance sheet.
The second axis of this mutation is embedded distribution, which consists in offering coverage at the precise moment the need arises, inside another purchase journey. Rather than drawing the customer toward an insurance brand, one inserts the insurance into a pre-existing transaction, credit, the purchase of a pet or general retail. This logic has given rise to partnerships where the insurer fades behind the platform that distributes its product, and where technology serves no longer to seduce directly but to integrate discreetly. Here again, the value of insurtech no longer lies in replacing the insurance chain, but in smoothing it. The share of the sector's deals backed by artificial intelligence has meanwhile jumped to around two thirds, confirming that capital is reorienting toward tools rather than brands10.
It would be unfair to conclude a pure and simple failure, for technology kept part of its promises, but on the front it truly mastered, that of operational efficiency. The automated handling of claims and underwriting genuinely reduced delays and costs, and enabled tangible margin gains. Lemonade, whose bots handle most routine requests in minutes, saw its gross margin rise from around 29% in 2023 to 39% by mid-2025, and its loss ratio improve from around 79% in 2024 to 67% the following mid-year, thanks to better-tooled pricing and underwriting6. Technology therefore did not fail, it simply proved its value at the front and at the counter, where it compresses friction, and not at the actuarial heart of the trade.
More revealing still, these gains did not remain the preserve of the insurtechs. The established insurers in turn adopted the automation of claims, data-enriched pricing models and digital journeys, often by acquiring the start-ups that had developed them or by forming partnerships with them. The diffusion of innovation thus went from the disruptors toward the incumbents, and not the reverse, which constitutes the sharpest rebuttal of the disruption thesis. The value created by insurtech was indeed real, but it was captured for a large part by those it claimed to oust, turning a promise of replacement into a program of modernization for the whole industry.
The recovery of the pioneers is itself instructive, for it was obtained not by moving away from classical insurance but by drawing closer to it. Hippo returned to a net profit of 58 million dollars in 2025, against a loss of 41 million the previous year, but its combined ratio, though much improved, remained at 113, a sign that its underwriting was still in deficit and that the profit owed partly to asset disposals5. Root slipped back into a slight loss after a profitable year, and Lemonade, despite a continuous reduction of its losses, was still in the red at the end of 20256. The path to profitability everywhere consisted in adopting the underwriting discipline, pricing prudence and cost management of traditional insurers, that is, in becoming a little more what insurtech claimed to surpass.
At the end of this trajectory, the initial question, was the promise of insurtech kept or betrayed, calls for a subtler answer than the alternative suggests. It was neither one nor the other, it was transformed. The promise of disruption, in the sense of a replacement of insurers by software companies, was indeed betrayed, because it rested on a diagnostic error. Insurance is not a software problem that technology could solve, it is a balance-sheet and risk-bearing problem that technology could at best lighten. The capital to immobilize, the solvency to respect and above all the underwriting to master form an irreducible core that no interface dissolves. It is this core that disciplined the disruptors and brought them back toward the trade.
But another promise, more modest and more durable, was kept. Technology reduced the cost and friction of insurance wherever the stake was not the bearing of risk itself, in distribution, claims handling and customer experience, and it supplied the established players with the tools of their own modernization. The frontier where insurtech creates value turned out to be the one where it serves the insurance function instead of claiming to supplant it. This lesson goes beyond insurtech, for it foreshadows the probable fate of artificial intelligence in insurance, the subject of other articles in this series, namely a powerful lever of efficiency and pricing, and not a substitute for the risk-bearing trade11. The real question of the coming years is therefore no longer whether technology will disrupt insurance, it will not in the sense once meant, but how far it will be able to compress its mechanisable part, and including pricing as close as possible to the policyholder's behavior, whose promises and excesses the next article in this series will explore.
This frontier of behavioral pricing is precisely the one where the lesson of insurtech will be tested. If technology cannot abolish the bearing of risk, it can on the other hand refine the knowledge of that risk down to the scale of the individual, by exploiting data on driving, health or usage. The promise is then of fairer pricing, rewarding virtuous behavior; the possible drift is that of an insurance which, by dint of individualizing risk, comes to destroy the pooling that founds it. Insurtech will have shown that technology does not replace the insurer; it remains to be seen whether, by making it too clear-sighted, it does not threaten the very idea of insurance.
The AI Act is a safety code, not a liability regime. Yet the withdrawal of the dedicated directive has made it, by default, the keystone of liability for artificial intelligence, and compliance with the regulation has become the new frontier of insurability.
Solvency II rests on two postulates inherited from classical actuarial science, the ability to diversify weakly correlated risks and the ability to estimate their distribution from historical data. Cyber risk contradicts both.
Since 2018, the financial lines market has offered the purest textbook case of the underwriting cycle. A cleansing of supply collided with the fear of a wave of Covid-related claims to produce the most brutal hard market of the century.
A leader makes a decision, a merger, a launch, a restructuring. It fails, the shareholders lose money and sue them, personally. Their house, their savings are on the line. And yet they sleep soundly, because an insurance exists that covers them against the consequences of their own judgment.
One analysis with every new publication, on insurance, reinsurance, cyber and AI. No account, no noise.
No spam, one-click unsubscribe, no data ever sold.