What happened

In a report published on 14 August 2026, Business Insider described Michael Burry’s concerns about enthusiasm surrounding AI. The investor compared the confidence he was seeing with the optimism that preceded the dot-com and housing market crashes. His comments appeared in subscriber discussions during that week.

Those comparisons are an investor’s interpretation. They do not demonstrate that AI is a bubble, establish when markets might turn or show that an individual company’s AI project lacks value.

For a business leader, the useful response is to examine the assumptions beneath the next spending decision.

Why it matters

A technology can be useful while a particular investment in it disappoints. A customer service assistant might improve response quality but cost more to operate than expected. A successful pilot might depend on a specialist who cannot supervise every department. A popular internal tool might save minutes without changing any customer outcome.

I would therefore separate three questions in an investment paper: does the capability work, does it improve this process, and does the improvement justify its full cost? Evidence for one question should not silently become evidence for the other two.

Consider an illustrative proposal to accelerate tender preparation. Faster drafting is a plausible benefit. The business case should also account for checking factual claims, maintaining approved source material, protecting confidential information and reviewing the final submission. Count accepted tenders and staff time released for useful work, alongside document production speed.

That produces a decision a finance director can challenge constructively.

The bigger shift

The more persuasive the industry narrative becomes, the more useful a modest, measurable commitment can be. A company does not need to resolve the future of AI to decide whether a bounded project deserves another month of funding.

Define an initial budget and a decision date. Record the existing cost and quality of the process. State what would justify expanding the project, changing it or stopping it. Include the cost of returning to the previous method.

A stress test should also ask what happens if supplier prices rise, demand is lower than expected or review takes twice as long. These are planning scenarios, not predictions. Their purpose is to expose a proposal that works only when every assumption is favourable.

This is where AI strategy should guide tool selection: the desired business result determines the investment, the sequence and the limits.

My take

I would not make a corporate AI decision by choosing between a famous sceptic and a confident technology founder. Both may draw attention to important questions. Neither knows the economics of a process inside your organisation without examining it.

Ask the team to present the strongest case against its own proposal. Which assumption contributes most to the expected return? What evidence would change its recommendation? Which expense has been left with another department?

Then approve the smallest useful test that can answer those questions. Protect worthwhile experiments, but give them an honest stopping rule.

A credible AI programme should survive a discussion about costs, alternatives and disappointing results. If its justification relies mainly on the fear of being left behind, the next investment should be in clarifying the business case.