Does the time AI saves at work actually benefit you or your employer?

AI Saves You Time, Your Employer Keeps It: The Sept 2026 Study

Mostly your employer, unless you decide otherwise. The UCD-led GenAI Use in Work report (Working in Ireland Survey, 4,300 workers, published September 7, 2026) found most people who save time with AI reinvest it in more work, a third say their pace has intensified, and only 4.7% earned more. Dorie Clark’s The Long Game supplies the fix: assign 20% of saved hours to white space and learning before they appear.

Every few weeks a study lands that says AI is saving workers time, and every few weeks I ask the same question in coaching sessions: saved for whom? On September 7, 2026, a team led by Professor John Geary and María Belizón of UCD with Chris Smart of the Nevin Economic Research Institute published the answer under the plain title GenAI Use in Work: Its Reach, Effects, Risks and Limits. It draws on the Working in Ireland Survey, 4,300 workers across the Republic and Northern Ireland, and it is the first dataset I have seen that follows the saved hour all the way to who ends up holding it.

The headline numbers are unglamorous and therefore believable. About a third of workers now use AI in their jobs, and almost half of those use it daily. Eighty-nine percent of users say it improved the quality of their output. Then the fork: most people who save time with AI say they reinvest it in taking on more work rather than in any better balance, a third report the pace of their work has intensified, and just 4.7% report any increase in earnings as a result. Only 1% want AI curtailed for fear of their jobs, which tells you the surplus is not being resisted. It is being quietly absorbed.

The 4.7% Ledger — the four-step protocol I run with clients

Log saved hours for two weeks. Every time AI compresses a task, write the task, the tool, and the honest minutes saved. My own ledger runs around three hours a week, not thirty, and the number matters more than the feeling.
Assign a destination before the hours exist. Dorie Clark’s rule in The Long Game is 20% time: one-fifth of what you reclaim goes to white space, learning, or an experiment with a seven-year horizon, booked in the calendar before the week starts.
Convert the rest into output that carries your name. The study’s 4.7% is what happens when saved time dissolves into throughput nobody can attribute. A shipped project, a decision memo, a client outcome is attributable; a faster inbox is not.
Run the pace check monthly. If your calendar is fuller than it was before AI, the refill has won. Cut one recurring commitment and put the hour back in step two.

The reason I reach for Clark here is that she diagnosed the mechanism years before the tools existed. Busyness, she writes, is often a mark of servitude, not importance, and you cannot pour more into a glass that is already full. AI empties part of the glass. The survey shows what most people do next: they pour more of the same work back in, which is precisely why the employer captures the gain. Clark’s Career Waves make the failure specific. Reaping without cycling back to learning is how you stagnate, and a saved hour spent on a fourth status report is reaping. The report has a second finding that sharpens this: nearly half of workers with a master’s degree or doctorate use AI at work against under 4% of those with a primary degree or secondary education, and 60% of the training people received lasted less than a day. The learning gap is where the compounding lives.

There is a more hopeful line in the same data. Around 60% of AI users say the reclaimed time goes to more complex or creative tasks, which is the outcome advocates promise. But the authors add the caveat that matters: a substantial share of those who now save time previously spent it on mundane work, and that is the lane the researchers flag as most exposed to displacement. If the only thing AI did for you was remove the boring part of a job that was mostly boring, the surplus is not yours to keep. The question a high performer should ask is not how many hours AI saves but whether the hours land on judgment, creation, or relationships, the three places Clark says compound.

The limits, named. These are self-reported savings from one island’s labour market, not measured output, and the 4.7% pay figure may lag because compensation adjusts annually while tools adopt weekly. Founders and solopreneurs capture the surplus directly and still fall into the refill trap, just with themselves as the employer. And an HRreview essay published September 8, 2026 makes the point the survey cannot: a minute saved by the person drafting is often a minute spent by the person checking, so at team level the gain may simply move. Track that too, or your ledger flatters you.

What I actually do is boring. A saved-hours column sits in my weekly review next to revenue, one reclaimed hour a week is a fixed learning block that no client call can take, and the pace check has cost me two recurring meetings this year. None of it requires a new tool. It requires deciding, before the hour appears, that it belongs to the version of you that exists in seven years and not to this week’s inbox.

Sources: The Irish Times on the GenAI Use in Work report (Sept 7, 2026), RTÉ — AI gains benefiting employers rather than staff (Sept 8, 2026), Working in Ireland Survey 2025 publication page, HRreview — the AI productivity paradox (Sept 8, 2026).