How to Stop Botsitting: Cut the Hidden Time Cost of AI

You've probably felt it: you ask an AI tool to draft something, and instead of saving time, you spend the next twenty minutes checking its numbers, rewriting the parts it got wrong, and feeding it context it should have already had. There's a name for this now: botsitting.

Glean's Work AI Institute surveyed 6,000 full-time digital workers across the US, UK and Australia for its Work AI Index 2026. It found that 87% of workers use AI and say it saves them 11 hours a week. But the same workers spend an average of 6.4 hours a week on "botsitting" — the work of making AI usable: feeding it missing context, checking its output, debugging its mistakes, rerunning prompts, and cleaning up confident-but-wrong answers. Only 13% say their organization is actually performing better because of it, a gap the CIO report on the study covers in more detail. Harvard Business Review's own piece on the study, How Much Time Do Your Employees Spend Botsitting?, by Rebecca Hinds and Paul Leonardi, puts it plainly: employees are spending nearly a day a week on this, and most of that labor is invisible and unrewarded.

If you use AI regularly for work, this isn't a reason to stop. It's a reason to get deliberate about where your time actually goes.

What botsitting looks like day to day

It's rarely one big task. It shows up as small, repeated interruptions:

• Re-explaining context you already gave the tool last week, because it doesn't remember your team's format, your client's preferences or last quarter's numbers.
• Line-by-line checking of a report, email or code snippet before you can trust it enough to send.
• Rerunning the same prompt three or four times because the first two outputs missed the point.
• Fixing confident errors — a wrong figure, a made-up citation, a policy detail that's out of date — that look right until you check them.

None of this is wasted time exactly. Some of it is necessary supervision. The problem is that almost nobody tracks it, budgets for it, or designs their workflow to reduce it.

The flip side: botshitting

The Work AI Index also names the opposite failure mode: "botshitting" — shipping AI-generated work you haven't actually verified. Around 69% of AI users admitted to doing this at least once, according to the same report. Cutting botsitting time isn't just about speed; unchecked, it turns into the kind of error that damages trust with a client, a manager or a regulator. The goal here is to spend less time babysitting AI in the wrong places, not to skip verification where it genuinely matters.

A practical framework to cut your botsitting time

Hinds and Leonardi's HBR piece points leaders toward three moves: identify when botsitting is actually necessary, build organizational context into the AI system so it stops asking for the same information, and measure quality and employee experience alongside speed. Here's how to apply that as an individual or a team this week.

1. Sort your AI tasks into three buckets

For a week, note every time you use AI for work and roughly how long you spend checking or fixing the output afterward. You'll typically find three patterns:

• Low-stakes, low-check: first drafts, brainstorming, formatting. Light verification is enough.
• High-stakes, necessary-check: anything client-facing, financial, or compliance-related. Full verification is non-negotiable — don't try to shortcut this.
• High-check, low-value: tasks where you're spending more time fixing the output than you would have spent doing it yourself. This is where botsitting is quietly costing you money.

Stop using AI for the third bucket, or fix the setup so it moves into the first two.

2. Build context once, reuse it everywhere

Most botsitting exists because the AI doesn't have context it needs and you re-supply it manually, every single time. Instead:

• Write down your team's standard format, tone and house style once, as a short brief.
• Save it as a custom instruction, project file or system prompt in whatever tool you use, so you're not retyping it.
• Keep a running file of recurring facts — your product names, your pricing, your org chart — that you paste in or attach rather than re-explain.

This single change removes a large share of the "feeding it context" category of botsitting.

3. Timebox the check, don't skip it

Decide upfront how long verification should take for a given task, based on its stakes. A five-minute internal Slack draft doesn't need the same scrutiny as a customer email quoting a refund policy. Writing the time limit down before you start keeps you from either over-checking low-stakes work or under-checking high-stakes work.

4. Track it, even roughly

You can't fix what you don't measure. A simple weekly note — "AI saved me time on X, cost me time on Y" — is enough for an individual. For a team, ask people to flag which tasks feel like they cost more time to supervise than they saved. That's the raw material for point 1 done at team scale, and it's the kind of signal Hinds and Leonardi say most organizations aren't collecting today.

Why this matters more for Indian teams

Many Indian workplaces are still in the early, enthusiastic phase of AI adoption — trying tools across writing, coding and customer support without a shared playbook for when to trust the output. That's exactly the environment where botsitting quietly eats the gains that got everyone excited about AI in the first place. Teams that build context once and check deliberately, rather than reflexively, get the 11 hours back instead of losing most of it to the 6.4.

Reducing botsitting isn't a productivity hack, it's a skill — knowing what to trust, what to verify, and how to give AI the context it needs so you stop repeating yourself. That's the kind of judgment our AI Generalist Certification is built to teach, alongside the practical AI skills covered in what an AI generalist actually does.