Markets have a habit of reaching for a single idea and applying it with brute force. Right now, that idea is AI disruption, and it is being applied with extraordinary bluntness to an enormously diverse set of businesses. Yet the argument doesn’t always stand up to scrutiny.
Take American Express, which sold off in February on concerns AI will displace white-collar workers and reduce consumer spending.1 If you follow the logic, every consumer-facing business should be under pressure. Walmart is trading close to an all-time high. Then look at insurance brokers, which slumped sharply on concerns about a new AI car insurance comparison app.2 Yet the institutional broker market, serving multinationals with complex, global policy requirements that only a handful of companies in the world can underwrite, bears no resemblance to the car insurance business model. The market has lumped them together.
Fat tails and false narratives
The same pattern is visible across sectors.
The MSCI World Software & Services Index was down 24 per cent in the first quarter of 2026. This is bear territory, even though many of the companies it tracks have continued reporting strong earnings. Financial data businesses, legal services providers and professional information platforms have also been caught in the downdraft, despite different competitive environments, different customer bases and different switching costs and barriers to entry.
These are examples of conflating what could happen in a tail scenario with what is likely in the base case. Some might suggest the range of possible outcomes has simply widened. I would push back on that. What has changed is that the tail – the probability attached to certain adverse outcomes – has fattened. Two things have legitimately shifted: the capital being spent on AI has grown significantly,3 and the quality of what these tools can do has improved materially. But neither of them, on their own, justifies treating every business that sells data, software or information as if the tail scenario is the base case.
There is also a fundamental question about adoption pace that the market appears to have skipped over entirely. If you are the chief financial officer or chief technology officer of a large company and someone said you could replace your enterprise resource planning system with an AI model, you could not do it overnight. Changing a core enterprise system takes years because that is how complex technology deployment works
Finding value in an AI world
The question not being asked clearly enough is: where will the value actually accrue in an AI-enabled world? Will it accrue to the AI model providers, whose margins may compress as the underlying models become commoditised and competition intensifies? Or will it accrue to businesses with decades of proprietary data, deep client relationships and switching costs that allow them to deploy AI for their own benefit and charge for the result?
Some of the businesses that have sold off most aggressively are those built around exactly the kind of proprietary, hard-to-replicate datasets that make them potential beneficiaries of AI. Businesses providing data and analysis to professionals have spent decades building information assets that cannot be quickly or cheaply replicated. The competitive moats are real. The market, for now, is treating them as though they are not.
This is not to dismiss AI disruption risk – we consciously avoid owning certain businesses we think are genuinely threatened. But a consumer-facing language learning app is not the same as an institutional-grade financial data provider. A generic software tool with limited switching costs is not the same as a mission-critical platform embedded in a regulated workflow.
A familiar pattern
Markets are occasionally gripped by ideas that are partly right and almost entirely overapplied. AI disruption is real. There will be casualties. But the current environment reflects a familiar pattern in which technological change has amplified a theme to the point where large groups of businesses are being assessed through the same lens, regardless of the differences between them. The result, in our view, is substantial mispricing across a wide range of businesses where some have decided the outcome without doing the analytical work.
For investors willing to make those distinctions carefully, to separate businesses that face real structural threat from those that have simply been caught in the narrative crossfire, that gap between sentiment and reality should not be seen as a problem. It is where opportunity lies.
Endnotes
- Bloomberg, Stocks Hit by AI-Disruption Fears as IBM Tumbles, February 2026
- 9fin, Insurance brokers in the firing line as AI fears spread, February 2026
- Bloomberg, Big Tech to spend $650 billion this year as AI race intensifies, February 2026
