The markets are getting it wrong on AI risk
Concerns that AI will make entire sectors obsolete are overblown. These are not companies encountering AI for the first time, yet markets appear to be pricing them as if adaptation is impossible, says Michael Clark
The repeated sell-offs triggered by new AI product launches suggest markets are struggling to distinguish between headlines and reality. Each release is treated as evidence that entire sectors from advertising to analytics to legal publishing, face imminent obsolescence. That assumption looks increasingly fragile.
Recent headlines around AI-driven sell-offs have hit companies such as Publicis and Relx particularly hard. The concern is that new AI tools will make these businesses obsolete. Possibly, but step back and the picture looks more nuanced.
Closer inspection shows these companies are not standing still. Publicis’s model has become increasingly platform-led, with technology, AI and data embedded across media, creative and client delivery. Relx has been deploying AI across legal and risk products for years, well before the current wave, employing around 12,000 technologists and spending close to $1.9bn a year on technology. These are not companies encountering AI for the first time, yet markets appear to be pricing them as if adaptation is impossible.
In doing so, investors risk losing sight of how businesses and the use of AI within them actually work. Periods of technological excitement have a habit of pulling attention away from fundamentals. How revenues are generated, where costs sit, how clients use products, and how organisations evolve. When those questions are sidelined, market reactions become driven more by narrative than by analysis.
Adaptation
Part of the problem is that investors continue to assess AI primarily through announcements and product launches rather than changes in underlying operations. Product launches are framed as revolutionary, and that framing is often taken at face value. In practice, demonstrations tend to show what is possible in controlled settings rather than what can be delivered consistently inside organisations. Turning technical capability into repeatable outcomes remains complex, slow and highly dependent on people.
This has encouraged a broader assumption that incumbent firms will struggle to adapt. That may prove true in some cases, but it is far from universal. Large professional services and analytics businesses have navigated multiple waves of automation before, reorganising workflows, repricing offerings and retraining staff along the way. Treating speed of innovation as evidence of inevitability risks overlooking the capacity of established organisations to evolve.
At the same time, markets appear to be underestimating where value resides. As raw outputs become cheaper and more abundant, differentiation shifts away from production and towards interpretation. Domain expertise, understanding and the ability to apply information to real decisions remain scarce. AI does not eliminate the need for judgment; it increases the premium placed on it.
A related error is the assumption that the average business or individual can immediately extract value from AI-generated data. In practice, most organisations struggle with interpretation, governance and integration. Access to information is not the same as the ability to use it well. Firms that help bridge that gap retain relevance even as AI improves.
Every AI release is treated as confirmation that existing business models are doomed, regardless of evidence
When these factors are ignored, market reactions become self-reinforcing. Every AI release is treated as confirmation that existing business models are doomed, regardless of evidence. This creates exaggerated price moves and encourages a narrative of inevitable displacement rather than gradual reconfiguration.
None of this implies that AI poses no risk. Margins will come under pressure, and some activities will be commoditised. But pricing every development as a zero-sum threat assumes both universal technological adoption and universal human competence, two conditions that rarely hold.
Markets need to move beyond reflexive reactions to headlines and return to fundamentals. The current approach to pricing AI is not sustainable and risks leading to either a loss of confidence or a bubble that eventually bursts. The focus must shift to who adapts, who retains pricing power, and who can translate capability into outcomes. Until that happens, AI announcements will continue to trigger volatility driven more by expectations than by reality.
Michael Clark is a market and industry technology adviser and author: www.michaelclark.ai/