92% of execs believe in productivity, 40% of staff see no time savings

AI Productivity

AI Productivity

92% of execs believe in productivity, 40% of staff see no time savings

3 min read

One of the most unpleasant stages of AI adoption begins at the moment when "acceleration" suddenly turns into additional work of fixing what AI itself generated.

That is why the topic of workslop looks important to me.

We are talking about content or work outputs created by AI that at first glance look high-quality, but in fact contain errors, contradictions, and require manual refinement. In other words, the company formally gets "acceleration," but in practice — a new layer of hidden load.

According to BetterUp Labs and Stanford Social Media Lab, 40% of office employees have already encountered this effect at work and on average spent about 3.4 hours per month fixing AI outputs.

That is exactly the uncomfortable moment that usually gets lost behind beautiful presentations about productivity gains.

The problem becomes even more noticeable when a gap appears between leadership perception and frontline reality. In one survey, 40% of rank-and-file employees said AI does not save them time at all, while 92% of top executives said AI makes them more productive.

So at the top, people often see potential. And at the bottom, they are cleaning up the consequences every day.

This is exactly where most overly optimistic rollout scenarios break.

If you cut headcount at the same time and demand "use AI more actively," the effect can be the opposite: drafts start appearing faster, but the final work takes more time because of rewriting, checking, and eliminating errors.

For me, this is one of the first truly applied markers of AI adoption maturity.

The question is no longer whether AI can generate something. The question is whether it reduces the total amount of work after generation.

If not, then it is not automation of the result, but automation of the error stream.

That is why when rolling out AI, it makes sense to measure not only the speed of the first answer, but also:

how much time went into fixes,
how many checks were added,
how many errors made it to the next stage,
and where exactly "acceleration" turns into hidden costs.

My position is simple: if AI adds cleanup work for the team, the problem is no longer employee discipline, but rollout quality, usage rules, and the control loop.
And in your practice, does AI actually save time — or does it more often just accelerate the appearance of drafts that then still take a long time to polish by hand?

Sergei Isakov | COO, Co-Founder