Your team saved 10 hours this week. Where did 3.7 of them go?

That pause is the whole issue. The time is real. The tools work. But the saving never turns up anywhere you can point to — not in the capacity plan, not in a resourcing conversation, not in anything you could show your own manager.

There is now decent data on where it goes. It leaks in two places. You can fix the first with better tools. The second is not a tools problem at all, and it is the bigger of the two.

One to seven hours a week, and that part is not in dispute

Workday, working with Hanover Research, surveyed 3,200 people — half leaders, half employees — across North America, Europe and Asia in November 2025, all of them at organisations with at least $100 million in revenue.

85% said AI saves them between one and seven hours a week. 77% reported higher productivity over the past year. Whatever else is true, the tools are doing something.

The problem starts immediately afterwards.

37% goes straight back into fixing the output

Workday calls it the AI tax on productivity: for every ten hours of efficiency gained, nearly four are lost correcting, clarifying or rewriting low-quality AI-generated content. Only 14% of employees consistently end up net positive once that rework is counted.

The burden is not spread evenly. The heaviest AI users take the worst of it — Workday found highly engaged employees lose an average of about a week and a half a year to rework alone. The people getting the most from the tools are also the people paying the most tax on them.

Workday and Hanover Research, 3,200 respondents, November 2025.

This is the leak you can do something about. Say clearly what "finished" means before work reaches a reviewer. Use tools built for the job instead of one general tool for everything. Both will bring the number down.

It is tempting to read that as a people problem. It isn't.

The obvious conclusion from a 37% rework rate is that people need training — better prompting, more discipline, a bit of coaching.

Workday reads it differently. They say the problem is the system, not the behaviour: AI was dropped on top of jobs that nobody redesigned around it. Work comes out faster, but the standards for accuracy, and who answers for it, never changed. Nearly nine in ten organisations say fewer than half their roles have been updated for AI.

Regular readers will have seen this shape before. In Issue 10 the equipment held and the procedure didn't. Here the output changed and the process around it didn't.

Nobody decided what the rest of the time was for

Of the organisations that do redirect freed-up time, Workday found 32% simply increase existing workload. 39% put it back into buying more technology. 30% into developing people.

The reason is simple. In a Gartner survey of 114 HR leaders, just 7% of organisations give any guidance on how to use time saved by AI. Small sample, so treat it as a guide rather than a precise figure. The direction is clear enough.

Where there are opinions, they do not match. 55% of those HR leaders would want a freed-up hour spent on projects outside the day job. In a separate Gartner survey of 1,973 managers, only 28% would choose the same. The people writing the policy and the people applying it want different things.

So the team works it out for themselves. In most places they conclude that finishing early just means being handed more work.

The redeployment decision

The three options below are ours, not theirs. The studies measure the problem. Neither suggests a fix. Pick one for each process rather than trying to do all three. What matters is that someone decides.

Say it out loud, then check it in 90 days

The decision only works if the team hears it. Two sentences in your next team meeting:

The last part matters most. Without it, people assume the opposite.

Then put a reminder in the diary for 90 days out with one question on it: did the time go where it was meant to, or did it turn into more work? If something went wrong in between, check whether the process covered it before asking who missed it.

Worth answering honestly before you decide what to change.

One caveat worth stating plainly

Both studies were run by companies that sell related services. Workday sells workforce software; Gartner sells advice. And the 7% figure comes from just 114 HR leaders. So treat the numbers as a guide, not as exact. They are still worth acting on, because the advice does not depend on them being right to the decimal. If the real figure were 20% rather than 7%, you would do the same thing.

One course on seeing where the time actually goes

It isn't about AI. It is about the gap between where you think the effort goes in a process and where it actually goes. You can watch the videos free; the assessments and certificate are paid. Content and pricing change, so check before you enrol.