How Much Time Do Your Employees Spend Botsitting? - Harvard Business Review

URL SCAN: How Much Time Do Your Employees Spend Botsitting? - Harvard Business Review
FIRST LINE: Summary.

The Dissection

The summary exposes a transfer, not a disappearance, of work. AI has not eliminated scheduling labor; it has relocated it into setup, supervision, exception handling, and error correction. Benjamin did not acquire a time-saving employee. He acquired a brittle machine requiring a human operator.

The Core Fallacy

If the article treats botsitting as evidence that AI cannot replace cognitive labor, it mistakes transitional friction for a structural limit. Under the Discontinuity Thesis, humans initially function as scaffolding around unreliable automation. That scaffolding is temporary if AI eventually produces competent output at lower total cost.

The real test is total labor cost, including supervision. An AI agent that requires as much human attention as the task itself is not automation; it is labor augmentation wearing an automation costume. But once the agent becomes reliable enough, the supervisory layer is compressed too.

Hidden Assumptions

• Employee supervision is treated as incidental rather than measurable labor.
• AI agents are assumed to have sufficient context, permissions, and reliability.
• “Time saved” is assumed to mean time returned, rather than labor silently reassigned.
• Human exception-handling is treated as permanent instead of a transitional layer.
• Management will measure total cost rather than celebrate the software deployment.
• The organization can preserve meaningful human participation while automation improves.

Social Function

Partial truth and transition management. The summary correctly reveals the hidden labor behind AI adoption, but it risks reducing a systemic transformation to an implementation nuisance. It warns managers about botsitting while leaving intact the comforting fiction that employees will remain necessary indefinitely as AI’s permanent handlers.

The Verdict

Botsitting is the corpse’s twitch of the old labor model. In the immediate term, it proves that many AI deployments are fraudulent cost accounting: the work has been moved into supervision, not removed. In the longer term, it is evidence of the opposite of human security. Employees are becoming servitors to machine capital, and once the machines no longer need constant babysitting, even the servitor layer becomes expendable. The decisive question is not whether botsitting exists today. It is whether its cost falls faster than human labor’s value. Under the Discontinuity Thesis, that race ends badly for the majority.