CEOs Who Think AI Replaces Their Employees Are Just Bad CEOs

A column published on Techdirt argues that executives who treat large language models as a substitute for their workforce are misreading both the technology and their own organizations. The author says four separate examples reached him in three months, each following the same script.

The all-hands pattern

In every case the trigger was an all-hands email in which the CEO praised LLM tools and told staff they must start using them immediately or look for a job elsewhere. Some companies brought in consultants to teach teams how to use the tools; others set up office hours or internal AI hackathons. The worst examples, the piece says, were companies that introduced token leaderboards — counting raw usage as a virtue, when good practice involves treating tokens as a scarce resource and a large share of usage is simply waste.

The column is explicit that it does not consider the tools overhyped in principle: used well, and chosen willingly, they can be powerful assistants. The willing part matters. Nobody forced into a tool learns to use it well, and a mandate produces compliance metrics rather than competence.

Levie's diagnosis

Box CEO Aaron Levie, himself an AI believer, is quoted on why executives are especially susceptible: they sit far enough from the last mile of work that generates most of the value with AI. Playing with the tools, they see happy-path results and skip the next ten or twenty steps needed for results that hold up. The prototype demo does not include reviewing the code before it reaches production; the generated contract does not include verifying the terms before they go to a counterparty. Levie's advice is the opposite of a mandate: use the tools heavily to understand what agents actually change in an enterprise, and come away appreciating both the upside and the work.

The column dislikes the phrase "AI psychosis" that has attached itself to this behavior, noting that psychologists and psychiatrists have called it inaccurate, but it agrees with the underlying observation.

Why the leap fails

The gap, the piece argues, is the distance between making something work and making it work well — well at scale, and well inside a specific environment. The employees a CEO cannot see are often the ones handling security, legal compliance or accessibility, which is why the leap from "I built a thing" to "anyone can build a thing" misses the point of hiring experienced people. The comparison offered is cargo-cult thinking: watching staff type and seeing work emerge is not the same as understanding what they do.

Layoffs as an excuse

Companies that cite LLMs to justify large layoffs are, in most cases, using the technology as a cover story: they over-hired, and "AI efficiencies" is easier to sell to Wall Street than "we made bad headcount decisions". The tools may let employees accomplish more; that does not mean fewer people are needed, only more people who know how to work productively with them. The conclusion is blunt: a CEO who believes AI replaces the work of employees is simply a bad CEO.