AI botsitting: Why your productivity gains aren't what they seem
AI promised productivity gains, but workers now spend hours correcting its errors with AI botsitting -- a hidden cost draining morale and ROI across organizations.
The botsitting crisis. Workers spend 6.4 hours weekly fixing AI mistakes -- more time than AI saves them -- reducing net productivity gains to just 4.6 hours per week and increasing turnover risk by 73% with frequent botsitters.
Root causes. Botsitting stems from a lack of enterprise context, premature deployment, integration failures, training gaps and misaligned incentives that reward AI usage over output quality.
Strategic actions for leaders. CIOs must prioritize quality over usage, build judgment capabilities, invest in unified AI platforms with organizational context, establish governance guardrails, and measure meaningful metrics beyond adoption rates.
Over the past few years, organizations have invested billions in AI, hoping to boost productivity and improve workflows. However, according to Glean's Work AI Index, AI may be harming productivity rather than helping it. According to the report, workers are spending 6.4 hours a week fixing AI's mistakes to make its outputs usable -- which is more time than workers are using AI for productivity.
The unrecognized and untracked labor and time that employees spend making AI usable -- including checking outputs, debugging mistakes, cleaning up incorrect answers and rerunning prompts, also known as "botsitting" -- is eating up nearly a full workday per week.
Botsitting isn't just a time-suck for employees -- it can actively hurt morale and retention. According to the Glean report, the more time workers spend botsitting, the more worn out they feel. Frequent botsitters are 73% more likely to be looking for new jobs. Unaddressed AI botsitting can create a ripple effect of consequences across organizations.
What is botsitting?
Botsitting includes any unmanaged work that workers do to make AI outputs accurate and usable. Botsitting can include several different activities that employees must perform, including:
Feeding AI missing context and organizational knowledge, such as company-specific products and internal processes.
Verifying AI outputs are accurate and identifying any hallucinations.
Switching between disconnected AI tools that don't share context.
According to Glean, workers save an average of 11 hours per week using AI, but after the 6.4 hours spent on botsitting, the net productivity gain from AI is only 4.6 hours per week.
As AI expanded, organizations expected AI to work autonomously. However, as AI got integrated into everyday workflows, the reality became clear: AI requires constant human supervision and correction. This causes a coordination neglect problem, where employees focus on individual productivity but underestimate the coordination required to translate these gains across teams or organizations.
The botsitting-to-botshifting cycle
As workers rely more heavily on AI, their engagement and accountability in their work decrease. Workers take more shortcuts with AI, and the bar for what is considered "good enough" in the final product is lowered, rather than iterating on AI responses to create the best-quality work.
Workers who are burnt out and frustrated with botsitting will offload more of their work to AI, and stop verifying sources or double-checking AI's outputs. According to Glean's report, 41% of workers have delivered AI-generated outputs they couldn't explain if asked, and 12% knowingly delivered output they believed was wrong.
This causes inaccurate or low-quality AI content to circulate through teams, integrate into workflows, and get passed on to customers. As low-quality information gets spread across the organization, it creates both internal and external risks to the organization's brand, reputation and more.
How CIOs are seeing botsitting play out in their organizations
Botsitting isn't just a point of frustration. It's also a productivity drain. The consequences of botsitting can cause ripple effects across the organization, showing up through gaps in ROI, training and adoption, metric tracking and governance.
For example, some workers may be stuck botsitting AI due to a skills gap. "Some individuals had very mature agentic systems set up, where they had invested time and energy to ensure an agent had all the context, connections, and instructions required to complete a task very effectively," Christian Chung, director of engineering at Fueled. "Others, in the same role and attempting the same task, but with a less mature agentic workflow, produced far inferior output and required far more human intervention to yield something usable."
"What makes it dangerous is that it doesn't look like a problem," said Chung. "On paper, adoption looks healthy: people are using the tools, and work is going out the door. But a meaningful share of the return you're paying for is quietly leaking back out as human cleanup time that never shows up in any metric."
Root causes of botsitting
Botsitting doesn't come from individual workers -- it's a systemic, organizational issue that can quickly become widespread as AI usage surges across the organization without oversight.
Botsitting can be caused by:
Lack of enterprise context. Since large language models (LLMs) do not have company-specific data or organizational knowledge to build off of, AI doesn't have context for specific products, customers, processes, or terminology. Instead, workers must feed AI the additional context needed for quality results.
Premature deployment. Many companies make the mistake of rushing to implement AI into workflows without the right preparation and oversight strategies. Leaders overlook pilot testing and iteration to appease boards and stakeholders, leading to a "deploy first, figure it out later" mentality that can slow workflows, waste money and create security risks.
Integration failures. The absence of an enterprise-wide AI strategy or a platform-specific approach can cause AI tools to operate in silos, leading to fragmented data across systems. Employees then must be "go-betweens" to move and sync data across systems that can't interoperate.
Training and change management gaps. When organizations rush into AI implementation, employees often miss critical, foundational learning opportunities, including proper training on AI use and prompt engineering, clear guidance on when AI should and shouldn't be used, education on AI risks and hallucinations, and frameworks for evaluating output.
Confident-but-wrong answers. AI can give convincing but incorrect answers and often gives users the answer they want to hear rather than the true one. Workers – especially those who spend hours a day reviewing AI outputs – can struggle to distinguish accurate from hallucinated content.
Misaligned incentives. Workers are incentivized and rewarded for using AI without any focus on the quality of their use, while those who spend significant time botsitting go unrecognized. Although AI usage might look good on the surface, it can mask hidden AI supervision costs and inefficiencies.
Strategic actions for CIOs
Once leaders identify the root cause of botsitting, they can create a strategy to address the core issues and reduce botsitting across the organization. "CIOs must treat AI like any other enterprise capability," said Ha Hoang, CIO at Commvault. "They need to start with processes that are well understood, define clear success metrics, establish governance, and continuously measure outcomes."
The goal shouldn't be to maximize AI usage; it should be to maximize business value while minimizing unnecessary human oversight.
Ha Hoang, CIO at Commvault
Focus on maximizing quality, not usage
Overusing AI without a clear strategy can lead to low-quality results at a high cost. To maximize quality, AI should be grounded in an enterprise context, not just data. Leaders should define clear guidance and standards for how, when and why AI should be used, including use case criteria, approval processes, output verification and quality assessment.
"The goal shouldn't be to maximize AI usage; it should be to maximize business value while minimizing unnecessary human oversight," said Hoang.
Build judgment capabilities
Judgment capabilities should be built across the organization before implementing AI, including training leaders and managers to evaluate AI outputs, developing coaching frameworks for responsible AI use, and encouraging employees to share AI tips and best practices.
Invest in integration and context
When making decisions about which platforms to purchase, leaders should find a unified process across the organization and consider strategic vendor consolidation to reduce AI tool sprawl. Additionally, enterprise-wide AI should have access to internal knowledge bases to improve context and ensure the technology maintains it across sessions.
"You need to work on a knowledge base for AI agents," said Yuri Gubin, CTO at DataArt. "The more you put into it in a way that is consumable by AI, the better. It should … be in the form of plain text that is straight to the point, factual and focused on the content."
Measure what matters
Organizations should choose valuable metrics that go beyond just usage, including output quality, botsitting time and patterns, and net productivity gains. In addition to examining how AI improves performance, leaders should assess employee satisfaction and retention for those who use AI, and create feedback loops and track data to identify where AI adds value and where it creates a burden.
Establish governance guardrails
Before implementing any AI, organizations should create AI usage policies with clear guidance on responsible use of AI, including examples of when human intervention is needed. Approval workflows should be established for any AI-generated customer-facing content.
Address the human side
Emphasize to employees the importance of AI output quality, not just its usage. Incentivize strategic, effective AI use. "Give them time, training and tokens so they can learn faster and more," said Mike Finley, CTO of AnswerRocket. "But in return, demand 'meta results' that are not just their own work but also lessons and policies you can spread around."
Be transparent with employees about how AI is being implemented and integrated into workflows and provide training and development for AI skills across the organization.
Alison Roller is a freelance writer with experience in tech, HR and marketing.
