Box CEO and Techdirt: AI mandates from the top reveal bad leadership, not bad workers

A wave of CEOs are issuing all-hands ultimatums demanding employees adopt LLM tools or quit, complete with mandatory hackathons, hired consultants, and — in the most misguided cases — token-usage leaderboards that reward burning compute rather than using it well. Techdirt’s Mike Masnick argues these mandates are self-defeating: AI tools genuinely boost productivity, but only when workers choose to adopt them and learn their limits. Coerced adoption produces neither.

Box CEO Aaron Levie, himself an AI optimist, diagnoses the root cause: executives are insulated from the ‘last mile’ of real work. When a CEO vibe-codes a prototype with an agentic tool like Claude Code, they see the happy path and conclude staff are redundant — never having to review the code, fix security issues, handle compliance, or do the dozens of unglamorous steps between demo and production. Masnick likens it to cargo cult thinking: pecking at a keyboard and seeing output isn’t the same as shipping safe, scalable products for real users.

The piece predicts companies betting on AI-driven mass layoffs will learn quickly that LLMs make good employees more productive rather than making them unnecessary — and notes that many ‘AI efficiency’ layoffs are really cover for earlier over-hiring, a story that plays better with Wall Street than admitting bad headcount decisions. Levie’s prescription: leaders should use AI heavily, precisely so they understand both its upside and the genuine work agents still require.