Refers to the ongoing human work required to make artificial intelligence (AI) systems produce reliable, accurate, and usable results. Botsitting includes providing missing context, refining prompts, monitoring AI behavior, verifying outputs, correcting errors, identifying hallucinations, rerunning requests, and integrating AI-generated content into real-world tasks. The concept draws attention to hidden labor that organizations often fail to recognize when estimating productivity gains from AI.

The term emerged in the mid-2020s as organizations increasingly adopted generative AI tools in everyday work. Researchers observed that while AI promised to automate routine tasks, employees often devoted substantial time to managing AI systems themselves. Activities such as supplying contextual information, checking factual accuracy, correcting misleading or incomplete responses, switching among multiple AI tools, and rewriting outputs became an increasingly common but largely invisible part of knowledge work.

Botsitting is both a technical and organizational activity:

• Technical: Reflects the current limitations of AI systems, which require human oversight because they lack contextual understanding, situational judgment, and the ability to consistently distinguish accurate information from plausible but incorrect responses.
• Organizational: Raises questions about workflow design, employee training, productivity measurement, trust, and psychological safety. When employees feel personally responsible for preventing AI errors or fear being blamed for mistakes generated by AI systems, botsitting can become time-consuming, stressful, and mentally exhausting.

As AI capabilities continue to improve, the nature of botsitting is expected to evolve rather than disappear. While routine verification tasks may decline, greater emphasis will be placed on supplying contextual intelligence, exercising human judgment, coordinating multiple AI agents, ensuring ethical and responsible use, and determining when human intervention is necessary.