AI agents save time but ‘botsitting’ erodes net gains in enterprises

AI agents are marketed as a way to accelerate office and specialist processes, yet a new risk class called “botsitting” – the ongoing supervision, error‑checking and manual correction of AI output – can significantly offset the promised productivity gains.

A survey of 6,000 full‑time employees shows that 87 % use AI tools, automating about 25 % of their work and saving roughly 11 hours per week. However, respondents also report losing an average of 6.4 hours per week to managing the AI and that 69 % of results are passed on unchecked, reducing the net benefit.

Experts stress that real value from AI agents depends on clear goals, robust data preparation, integration with existing tools and strong governance. Human oversight, defined approval loops and safeguards against hallucinations are essential to prevent the oversight burden from overwhelming the time savings and to turn AI agents into a sustainable competitive advantage.