Business Idiot

"Business Idiot" is Edward Zitron's term for a structural type, not a moral judgment of any individual. It names the C-suite and senior-management role whose daily output consists mostly of meetings, emails, lunches, and decision-signoffs on work other people have done — and whose performance can't be measured against an artifact they personally produced. Zitron's argument: a person in this role spends most of their working day producing an impression of work, and AI is built to produce exactly that.

The original framing is in Zitron's "The Era of the Business Idiot." It returns in ai-too-expensive as the structural answer to the question "why is so much enterprise AI spend happening with so little measurable return?"

The mechanism

Three features of the role make Business Idiots structurally receptive to generative AI:

• The output is performance. Slide decks, status updates, PRDs, demos to other executives. These are exactly what LLMs can fluently produce.
• The signal is upward. Whether your work is good is judged by people one or two layers above you, who are evaluating the appearance of competence, not running your output through tests.
• The model never says no. A regular engineer pushes back on a request that's unreasonable or impossible. An LLM says "of course, right away!" and burns whatever tokens are required. For an executive, this is the ideal subordinate.

Zitron's harsher framing: "Generative AI is really good at doing an impression of work, much like most managers and c-suite executives, and even if it's completely incapable of doing something, it'll absolutely say it can and tell you you're amazing for suggesting it."

Why this matters for the bubble argument

The Business Idiot frame is what closes the gap in Zitron's broader subprime AI thesis: someone has to be paying these enterprise prices, and the math of customer ROI doesn't work. Zitron's answer is that the buyers are not running ROI math. They are running social proof and FOMO math. Workato's CIO openly described the approach as "eating the costs while employees experiment."

The fact that AI vendors don't offer per-task cost transparency or SLAs (see ai-too-expensive on Anthropic's deliberate opacity) only works because Business Idiots aren't asking for them.

What this predicts

Zitron's structural prediction is that once the first Business Idiot cuts their AI budget — once cost-containment becomes the visible move — the rest follow, fast. The behaviour is herd-shaped on the way in; it will be herd-shaped on the way out. That moment is one of the pale-horse signals worth watching.

A second prediction is that the public AI-Native Engineering rhetoric (engineers as "soloists → conductors → composers," "never open a codebase again") collapses on contact with the production engineering reality. Zillow's internal slide deck and the 50% increase in human review work documented in ai-too-expensive are the worked example.

Related concepts

• ceo-ai-psychosis — HandyAI's framing for the same demographic with a focus on agent-orchestration grandiosity
• ai-sycophancy-loop — the Stanford-documented mechanism that makes LLMs especially effective on this user class
• tokenmaxxing — the inverted-employee version: workers gaming token-leaderboards because executives turned token usage into a status metric
• marius-rise-of-the-bullshittery — the broader political-economy account that names this dynamic at industry scale
• ai-great-leap-forward — the corporate-mandate version of the same disconnect
• average-is-all-you-need — the floor-raising side; Business Idiots are the customers most happy with average
• skill-atrophy-supervision-paradox — what happens to actual engineers under Business Idiot management