Nikhil Suresh has argued that AI mania decision making is being distorted across corporate life because executives have incentives to repeat implausible claims rather than challenge them. His essay, highlighted by Daring Fireball, centers on a private conversation with an unnamed Fortune 500 executive who described why skepticism about generative AI can be costly inside enterprise sales.
Suresh said he met the executive during an overseas trip and found them technically competent, despite their company having publicly committed to familiar claims about dramatic AI-driven productivity gains. In private, he asked whether the rhetoric was just sales material.
According to Suresh, the answer was only partly yes. The executive told him the stronger pressure came from customers. If a vendor executive publicly said that a customer’s claimed 100x productivity improvement was unrealistic, that could make the customer’s own leadership look foolish, be treated as an attack, and risk the loss of an enterprise contract. Suresh’s point is less about one company’s slide deck and more about the incentive machine around it: once buyers and sellers both benefit from the story, honesty becomes a career hazard.
What is AI mania doing to corporate decision-making?
Suresh’s argument is that AI enthusiasm has hardened into a corporate taboo against dissent. Daring Fireball summarized the essay as a claim that the whole corporate world, beyond the tech industry, is caught in a form of AI fervor that punishes heresy and pushes skeptics into silence.
That is a useful distinction. Generative AI tools such as ChatGPT and Claude are useful, and the commentary does not deny that. The claim is that some executives are treating present-day systems as if they have already delivered productivity jumps far beyond what is plausible, then forcing vendors and colleagues to behave as if those claims are settled fact.
Daring Fireball framed the wider psychology this way: computers have changed the world, but many people cannot use them as creative tools because they do not understand how they work. Generative AI changes the feel of that relationship. People without much technical ability can now ask a chatbot to draft, summarize, code, or explore topics that previously sat outside their reach.
That can feel like magic, including to people who do understand computers. For some confident corporate managers, the commentary argues, it feels like a first encounter with computers as tools for making things rather than merely communicating. The result is overcorrection: genuine utility gets inflated into claims that are several orders of magnitude larger than the technology can currently support.
The enterprise risk is plain enough. If buyers announce impossible gains, vendors may decide that correcting the record threatens revenue. If employees see skepticism punished, they stop reporting what they know. Suresh’s essay describes a market where the safest answer is not the most accurate one, which is a lousy way to buy software and a worse way to run a company.
This story draws on original reporting from Daring Fireball.
