Almost everyone is using AI at work. Far fewer are seeing it show up in results. That gap is AI’s hidden cost, and it has very little to do with the technology. It is a habits problem, and it is playing out quietly inside most teams right now.

The Productivity Paradox Nobody Budgeted For

Here is the pattern that should stop every leader in their tracks.

87% of digital workers now use AI at work. 75% say it makes them more productive, saving them roughly 11 hours a week through automation alone. Yet only 13% of organisations say they are performing significantly better as a result.

So where do the hours go?

According to the Work AI Index, a study of 6,000 workers by the Work AI Institute and Glean, they disappear into two habits worth naming. The report calls them BotSitting and BotShitting, and together they explain much of what economists now call the AI productivity paradox.

The First Hidden Cost: BotSitting

Botsitting is the hidden work of making AI usable. Feeding it context. Checking its output. Debugging it. Cleaning up after it. For every hour we spend getting something useful out of AI, we spend roughly another hour making it usable.

It looks like diligence. It quietly eats the time AI was meant to give back.

The numbers make it concrete. Workers now spend an average of 6.4 hours a week botsitting, which is more time than they spend actually producing work with AI. Feeding context is the single biggest drain, and debugging is the most exhausting of the lot.

There is a people cost here, not only a time cost. When this labour stays unrecognised and unrewarded, workers wear down. Frequent botsitters are 73% more likely to be actively hunting for another job.

The Second Hidden Cost: BotShitting

When people get worn out from babysitting the bot, they swing to the opposite extreme. They stop checking. They pass on AI work they have not verified and could not explain if asked.

The report calls this botshitting, and 69% of AI users admit to it. 41% say they sometimes hand over work they could not stand behind.

Here is the counterintuitive part. The more capable the tool, the worse this tends to get. The tools that deliver the biggest reported productivity gains also carry the highest rates of unchecked work. When output looks polished, the small friction cues that used to make us slow down and review vanish, so people stop looking closely.

Where AI’s Hidden Cost Really Comes From

Set the two habits side by side. One over-supervises everything. The other trusts everything. Both drain the value before it ever reaches the organisation.

This is why buying more licences rarely moves the needle. Access is easy to purchase. Judgement has to be built.

The space between adoption and enablement is exactly where AI’s hidden cost lives, and it is where the 13% pull ahead. They do more than use AI heavily. They build what the report calls the human infrastructure around it: clear context, quality standards, and the judgement to know when to trust the tool, when to check it, and when to leave it out altogether.

That final skill deserves more attention than most training gives it. Knowing when not to use AI is the hardest capability to build, and only about a third of workers feel confident they have it.

Ending thought:

AI’s hidden cost is a signal about the habits forming around the technology, not a flaw in the technology itself. The teams stuck among the 87% keep paying the bill in botsitting, botshitting, and a slow erosion of trust. The teams moving into the 13% are building the judgement that turns individual speed into organisational results.

If your organisation is feeling the gap between AI activity and real impact, this is exactly the kind of challenge Radiant Institute helps with. Reach out to explore how our AI enablement programmes can help your teams turn adoption into capability, and capability into results.