The Rise of Botsitting: How AI Supervision is Impacting Worker Productivity and Job Satisfaction

AI Supervision is Creating a New Class of "Botsitting" Labor

Workers are increasingly spending significant portions of their workweek—averaging over six hours—engaged in "botsitting," the act of supervising, correcting, and managing AI agents. While AI is marketed as a tool for productivity, the reality for many employees is a shift from performing skilled work to managing the errors and idiosyncrasies of automated systems.

This transition often replaces high-value, engaging tasks with low-reward administrative oversight. For example, customer service representatives who previously found meaning in building human relationships are now frequently tasked with supervising AI agents that handle those interactions, leading to a decrease in job satisfaction and a perceived loss of professional purpose.

The Productivity Paradox: Output vs. Effort

There is a growing disconnect between how management and employees perceive AI-driven productivity. This divergence is often characterized by two different metrics of success:

• Management Perspective: Productivity is measured by total output per hour of paid labor. If total output remains stagnant or drops slightly despite AI integration, it is viewed as a productivity loss.
• Worker Perspective: Productivity is measured by output relative to the effort exerted. Workers may feel more productive if they can achieve 80% of their previous output while only exerting 40% of the effort.

Furthermore, some users report that AI can create a "false sense of productivity," where companies focus on implementing new tools—such as MCP tools or agentic workflows—rather than focusing on actual productive outcomes, mirroring the way individuals once spent excessive time tweaking Notion templates or Pomodoro timers.

Technical Challenges and the "Botsitting" Workflow

For technical professionals, such as software engineers and sysadmins, the experience of botsitting varies based on the tooling and the nature of the work.

The Friction of Agentic Workflows

Some developers report that AI agents break the traditional rapid feedback loop of "write code, run, change, repeat." Instead, they must wait for agents to plan and execute, creating 5-10 minute windows of inactivity that sap patience and disrupt the flow of development.

Context Switching and Management Overhead

As AI allows for the simultaneous management of multiple tasks, the burden shifts to high-level context switching. One developer noted that using orchestrator agents to spin up worker agents results in a backlog of draft Pull Requests (PRs) that require review, making the process of recovering context for each task paralyzing.

Mitigation Strategies

Experienced AI users suggest that the frustration of botsitting can be reduced through specific technical guardrails:

• Sandboxing: Running agents in sandboxes to prevent tedious permission requests.
• Tooling: Using strict linting or validation tools (e.g., forbidigo for Go) to keep the AI "honest."
• Clear Guardrails: Setting strict boundaries to minimize the "bizarre" outputs that often follow streaks of brilliance.

The Psychological and Social Impact of Automation

Beyond efficiency, the shift toward AI supervision is creating significant psychological strain across various industries.

Loss of Craftsmanship

Many professionals express a sense of loss regarding the "craft" of their work. Software developers compare the current AI shift to the era when library management replaced writing bespoke functions; while the latter was more efficient, it reduced the joy of creation. This shift is described as a "sudden shock" to the system rather than a gradual transition.

Alienation and Job Security

There is a growing sentiment of social alienation as workers realize they are becoming "babysitters" for a workhorse. This has led to several critical observations:

"I theorise that many social ills come from workers having less pride in their skills and achievements, and a greater sense of social alienation, due to automation."

In creative fields like design and animation, the impact is more severe, with some reporting a sharp decline in job satisfaction (e.g., from 80% to 20%) as they are treated as "sub-human" by stakeholders who brag about replacing human designers with AI.

The Incentive Problem

Some critics argue that the current state of AI inefficiency is an incentive problem. Because AI providers often charge per token, there is little financial incentive for the models to be as concise or efficient as possible, potentially leading to more verbose or circuitous outputs that require more human supervision.