Your team isn't actually becoming twice as fast. They're just switching from one type of low-value work to another.
Silicon Valley promised that generative artificial intelligence would wipe out mundane tasks, clearing the way for deep, creative strategic thinking. But the reality on the ground looks radically different. Employees are pocketing massive hours of theoretical savings, only to dump half of that time right back into a frustrating new dynamic called "botsitting."
A massive study of 6,000 digital workers across the US, UK, and Australia by Glean's Work AI Institute reveals a bizarre corporate friction point. While AI automation saves workers roughly 11 hours a week, they waste an average of 6.4 hours a week managing, fixing, and cleaning up after the software. That's nearly a full day of every single work week spent acting as an uncredited babysitter for chatbots.
This explains why 75% of individuals state that AI makes them more productive, yet only 13% of organizations report significant business performance gains. The personal time savings are completely real, but those gains evaporate before they ever hit the corporate bottom line.
The Anatomy of the Botsitting Burden
Botsitting isn't just one single task. It's a collection of micro-frustrations that chip away at the working day. Workers aren't spending their time thinking big thoughts. They're stuck manually dragging the AI across the finish line.
The time lost to botsitting generally breaks down into distinct corporate chores:
- Feeding the context window (2.3 hours/week): Because enterprise data is messy and scattered, workers must manually collect information and upload it to the prompt. Doing this over and over causes severe fatigue. Researchers call this "context rot" because the more unorganized data you shove into a prompt, the worse the output tends to get.
- Supervising and verifying outputs (2.2 hours/week): AI generates text and data that look beautifully formatted and highly polished on the surface but are completely incorrect, incomplete, or missing nuance under scrutiny.
- Debugging and prompt engineering (1.7 hours/week): When an AI tool fails, employees have to play tech detective. They change the phrasing, swap models, inject more context, and rerun the same prompt until the system finally returns something usable.
- Tool sprawl and cleanup (0.2 hours/week): A staggering 77% of workers juggle multiple AI tools each week, and 33% use four or more. Because these apps don't talk to each other, employees act as human middleware, copy-pasting code, data, and text across different tabs.
The Slippery Slope to Botshitting
When workers burn almost a full day every week verifying AI work without any formal recognition, a predictable behavioral shift occurs. They burn out. They get tired of fighting the machine, and they start cutting corners.
This gives rise to an even more dangerous workplace trend: "botshitting."
This happens when an employee ships an AI-generated product that they haven't verified, don't fully understand, or couldn't defend if a manager asked them to explain the logic. The numbers should alarm any executive team. Currently, 69% of AI users admit to botshitting at work. Furthermore, 41% openly confess that they occasionally turn in work they couldn't explain if put on the spot, while 28% have used the AI as a scapegoat, blaming the bot for mistakes they actually made themselves.
It creates a terrifying structural risk for corporations, especially in fields like finance, healthcare, or legal services. Highly polished, professional-looking garbage gets passed up the food chain because the human layer of quality control is too exhausted to read the fine print.
High Achievers Are Actively Hiding Their Workflows
The lack of organizational structure has forced top tier workers underground. High AI achievers—the people who successfully use these tools to elevate their actual work quality—aren't sharing their secrets with leadership. Instead, they're running a shadow operation.
The data shows that 54% of these top performers use completely unapproved tools or use approved platforms in noncompliant ways. They don't trust the basic software provided by corporate IT because it lacks the necessary data context.
Even wilder? 36% of these high performers actively hide how much AI helps them, and 38% downplay its assistance to their managers. They realize that if they reveal they saved 15 hours on a project, their reward won't be a promotion; it'll just be double the workload. They choose to keep the extra time for themselves, using it to pace their day or avoid burnout.
Why Giving Workers More Tools Fails
Most executives think the answer to low AI output is buying a better, shiny new standalone agent tool. That's a massive mistake. Uber reportedly burned through its entire annual AI budget in the first four months of the year without shipping a single usable feature, proving that throwing money at unintegrated tools solves nothing.
The core issue is that 53% of workers say their AI tools don't have access to the internal data they need to do their jobs. When a tool is context-poor, employees are forced to spend their time feeding it data manually.
Organizations that are winning this battle aren't deploying more bots; they're radically redesigning how work happens. In companies that are successfully transforming, 90% of workers say their employer treats AI as an opportunity to change the actual workflow, not just speed up old tasks.
How to Fix Your Broken AI Strategy
If you want to stop paying the hidden tax of botsitting and protect your business from unverified bot slop, you need to stop focusing on the technology and start focusing on the human workflow.
Connect Your Data Sources
Stop making your employees act as the integration layer. If your AI tool forces a worker to copy data out of an email, paste it into a scratchpad, and upload it into a browser tab, you've failed. Invest heavily in central enterprise search architecture that connects your communication channels, cloud storage, and project management hubs directly to the AI model securely.
Redefine Performance Metrics
If you evaluate employees purely on volume, you are begging them to turn in unverified bot outputs. Shift your management focus from output quantity to output verification and strategic alignment. Acknowledge that reviewing and editing AI work is a real, exhausting skill that takes time.
Legalize the Shadow Workflows
Stop punishing your top tech-savvy employees with outdated compliance rules that block useful platforms. Interview your high achievers. Find out which unapproved tools they're using to get the job done, and build a secure corporate path to officially license and audit those tools.
Audit Your Tool Failure Rates
Track how often your employees run a prompt only to abandon it. With 36% of workplace AI sessions currently failing completely, your team is wasting hours on broken loops. If a vendor's platform requires multiple rewrites or constant model-switching to produce standard work, cut the contract. Reliability matters far more than a long list of features.
