How AI 'Botsitting' is eating into employee productivity
AI can save time, but “AI botsitting” is creating a hidden cost as workers spend hours guiding AI, checking its work and fixing mistakes.
AI was supposed to make the workday shorter. Instead, employees are finding themselves spending hours managing the tools that are meant to make them faster.
This hidden work is being described as “AI botsitting”, the time workers spend making AI usable by adding context, correcting weak responses, checking facts, debugging workflows and cleaning up AI-generated output before it can be shared.
As companies bring generative AI and AI agents deeper into everyday work, that extra layer of supervision could be quietly eating into the productivity gains promised by automation.
The hidden work behind AI productivity
A Work AI Index report released by Glean’s Work AI Institute on June 10, 2026 found that digital workers spend 6.4 hours a week on botsitting. The study surveyed 6,000 full-time digital workers across the US, UK and Australia, alongside workplace AI usage data.
It found that 37% of the time employees spend with AI goes into botsitting, slightly more than the 36% spent using AI to produce actual work. That creates an interesting gap.
While 87% of digital workers use AI at work and 75% say it makes them more productive, only 13% said AI had significantly improved their organisation’s performance. In short, an employee may finish an individual task faster, but the organisation still has to pay for the time spent checking, correcting and coordinating AI output.
Glean, led by founder and CEO Arvind Jain, links much of this problem to AI tools lacking business context. Employees may have to enter the same information into different applications, compare responses from multiple tools or catch answers that sound confident but are simply wrong.
That becomes more important in areas such as finance, legal, healthcare and customer support, where an unchecked AI response can create compliance or reputational problems.
AI adoption is not the same as AI productivity
The problem goes beyond any single AI tool. Gartner has warned that growing interest in AI agents is accompanied by concerns around governance, trust, hallucination protection and whether organisations are ready to deploy them effectively.
Office productivity is also often the default AI use case when companies have not clearly identified the business problems they want the technology to solve. For India, where enterprises, startups and IT services companies are rapidly experimenting with AI-enabled workflows, that distinction matters.
Simply measuring how many employees use AI, how many licences are purchased or how many prompts are generated will not show the full productivity picture. Firms may also need to measure the time employees spend reviewing AI output, switching between tools, correcting mistakes and supplying missing context.
Better access to company data, clearer review standards, employee training and stronger governance could help reduce that burden. AI can still deliver meaningful productivity gains, particularly for repetitive and information-heavy work.
But if companies ignore the human effort required to supervise AI, some of those gains could disappear into a new form of digital busywork.