Botsitting: Employees Lose More than 6 Hours per Week Fixing AI Errors

The promise of artificial intelligence has always been simple and alluring: automate repetitive tasks, free up time, and increase productivity. However, a new study conducted by the American company Glean sheds uncomfortable light on the corporate reality of using these tools — and the numbers reveal a troubling paradox. Instead of simply benefiting from automation, professionals are spending a significant portion of their work time supervising, validating, and correcting the errors of the very AI tools that are supposed to make their lives easier.

According to the research, published by TechRadar, employees in the UK spend an average of 6.3 hours per week on monitoring and correcting artificial intelligence platforms. This figure, on its own, is striking. But it becomes even more revealing when contrasted with another number from the same study: workers save, on average, 12 hours per week thanks to process automation. In other words, half of the time gained with AI is immediately returned to the task of managing the AI itself.

---

What is "Botsitting" and Why Does It Matter

The market already has a name for this phenomenon: botsitting. The term, a fusion of the words "bot" and "babysitting," describes exactly what happens in the corporate daily grind: professionals who spend a significant part of their workday watching over automated systems, ensuring that the generated responses are accurate, consistent, and usable.

Far from being a mere semantic curiosity, botsitting represents a real and growing operational cost for companies. It doesn’t show up in efficiency reports, isn’t accounted for in technology ROI metrics, and is rarely discussed in leadership meetings — but it’s there, consuming productive hours and silently redefining the roles of millions of workers.

---

High Adoption, Low Institutional Trust

The data from Glean's survey also reveals a landscape of massive adoption of AI tools in the British corporate environment. An impressive 90% of tech professionals in the UK report using artificial intelligence at work — a rate higher than that recorded in the United States, where the figure stands at 84%.

Despite this high level of adoption, only 42% of companies are classified by their own employees as truly AI-focused organizations. This indicates that, in practice, the use of tools remains fragmented, individual, and often disconnected from a coherent corporate strategy.

This dissonance deepens when we analyze the perception of impact:

• 78% of respondents believe that AI increases individual productivity
• Only 18% believe it has significant impacts on the overall performance of the organization

The gap between these two numbers is striking and symptomatic. Professionals feel the benefits in their own work but do not perceive — or trust — that this gain translates into real results for the business as a whole.

---

The Hidden Cost: Reviewing is More Common Than Creating

One of the most revealing findings of the study concerns the distribution of time that employees dedicate to AI tools. Historically, it was thought that the most significant effort would be in creating prompts — the initial commands and requests sent to the platforms. But the reality has reversed.

Today, professionals dedicate:

• 38% of their time with AI to reviewing and ensuring quality of the generated responses
• 36% of their time to creating initial commands and requests

In other words, checking if the AI made a mistake has become the primary activity related to the use of these tools in the corporate environment — even more so than the actual creative interaction with the systems.

---

One in Three Uses Completely Fails

Glean went even further in its data on reliability: more than one-third of sessions conducted with artificial intelligence tools fail completely. These are not imperfect responses that need minor adjustments — they are complete failures, resulting in substantial rework.

The practical impacts of this instability are felt by a large portion of workers:

• 77% of professionals had to redo or correct texts, code, or analyses generated by AI in the last month
• 26% faced the same situation in the last week

These numbers reveal that the rework caused by AI failures is not an occasional exception — it is a routine and frequent experience for the majority of users of these technologies in the corporate sector.

---

The Critique of Vanity Metrics

Rebecca Hinds, director of the Work AI Institute, does not hold back when analyzing how companies measure the success of their artificial intelligence initiatives. She argues that most corporations evaluate technology efficiency in a superficial manner, relying on what she refers to as "vanity metrics": volume of platform access, number of commands sent, number of active users.

These metrics, while easy to present in executive meeting slides, do not capture the invisible work that underpins — and often limits — the actual results of automation. The effort of supervision, validation, and correction performed by employees simply does not show up in productivity dashboards.

The expert's warning is a call for organizations to rethink their success metrics and adopt a more honest and comprehensive view of what it truly means to integrate artificial intelligence into the workplace.

---

What Companies Need to Do Differently

In light of this scenario, it is clear that simply adopting AI tools is not enough to ensure real productivity gains. Organizations that want to reap the promised benefits of intelligent automation will need to invest in several strategic fronts:

1. Qualified training for employees to use AI tools more efficiently, reducing errors from improper use
2. Careful selection of platforms, prioritizing solutions with higher reliability and lower error rates
3. Review of evaluation metrics, incorporating indicators that capture supervision time and rework
4. Strategic integration of AI, aligning individual tool usage with collective objectives and well-defined organizational processes
5. Management of invisible work, recognizing and accounting for the effort of botsitting as part of the operational cost of automation

---

Is AI Still Worth It? The Answer is Yes — With Caveats

The data from Glean's survey does not imply that artificial intelligence is useless or that companies should abandon their automation initiatives. After all, 12 hours saved per week remains a significant gain — even if half of that time is reabsorbed by botsitting.

The central point is that uncritical and unstructured adoption of AI creates an illusion of productivity. Organizations celebrate adoption numbers and individual reports of efficiency without seeing the cost being incurred behind the scenes, hour by hour, screen by screen.

For simple and well-defined tasks — text reviewing, data organization, draft generation, research support — AI remains a powerful and legitimate ally. The issue lies in treating it as an infallible and autonomous solution, when in practice it still requires an attentive and available human copilot.

Artificial intelligence is here to stay. But maturity in using these tools is still being built — and the first step is to look honestly at the numbers that challenge the narrative of frictionless progress.