‹ back to writing
Nov 17, 2026·6 min read
workai-at-work

Who gets the time AI saves at work?

AI saves time at work, but nobody decides where that time goes, so it usually becomes more work. In Dayforce's 2026 survey of 6,909 people, 64% want saved time used to cut workload and burnout, but 40% expect more work and 39% expect slower hiring or cuts. The fix is to decide, before you automate a task, what the freed hour is for.

I hand a lot of my week to agents now. One writes the first draft of an essay like this every three days. One fills a daily post slot. One reviews every night what broke and what's waiting on me.

This week I reread the instructions I gave them. Every routine says exactly what the agent should do, when, and what it must never touch.

Not one of them says what I should do with the hour it gives back.

my routines · what each one says
essay routine · what to do → written
essay routine · when → written
essay routine · never touch → written
essay routine · the saved hour → —
daily post routine · the saved hour → —
// Every rule was about the agent's time. None was about mine.

That is the whole essay. The time AI saves at work goes to whoever decides what it's for, and if nobody decides, it becomes more work by default. Workers mostly want saved time spent on less burnout and on learning. Most expect it to be filled with more tasks or fewer colleagues. If you want a different result, name the hour before you automate the task.

Where does the time AI saves at work actually go?

Mostly into more of the same work, if you ask the people doing it.

Dayforce's 17th Annual Pulse of Talent surveyed 6,909 workers, managers and executives at organisations with more than 100 employees across Australia, Canada, Germany, New Zealand, the UK and the US, online from 27 July to 14 August 2026, with Hanover Research. It asked what people want done with AI-saved time, and what they expect.

AI-saved time: wanted vs expected · Dayforce Pulse of Talent, 2026
  • Want it used to reduce workload and burnout
    0%
  • Want it invested in learning and development
    0%
  • Expect employers to give them more work
    0%
  • Expect reduced headcount or slower hiring
    0%
  • Expect it to improve work quality
    0%
// People know what they'd do with the hour. Most don't expect to get it.

The gap between the first two rows and the last three is the whole problem. Nobody is lying about the productivity gain. The disagreement is about who it belongs to.

What would workers do with the time AI frees up?

Mostly better work, not less work.

In November 2026, Mercor published What workers want to hand to AI, and what they want to keep, a survey of workers mapped to their occupation's tasks (built on the US Department of Labor's O*NET task structure). Asked what they'd do with time saved by AI, 38% said they'd do more of the tasks they enjoy, 22% said the same work at higher volume, 18% said move into new kinds of work, and only 15% said work fewer hours.

The same survey shows which tasks people want to give away and which they want to keep. They want to hand over clerical work, like keeping records up to date. They want to keep the work built on human interaction and the high-stakes calls: for software developers, mentoring juniors and reviewing designs; for lawyers, interviewing clients and advising them on risk.

Saved time doesn't have a default owner. If you don't give it one, the backlog takes it.

Why does AI-saved time turn into more work?

Because the backlog is always closer than the plan.

Dayforce's own numbers explain it. 60% of respondents say understaffing has become normalised. 50% say they're expected to build new skills while keeping the same workload, but only 27% say they have enough time at work to do it. Executives see it differently: 56% of them say employees do have the time.

On a thin team, a freed hour doesn't sit there waiting for a decision. It gets eaten by whatever was already late. That isn't anyone being cruel. It's what happens when the automation was planned in detail and the hour it produces wasn't planned at all.

Dayforce put it plainly in a follow-up post: if AI removes five hours of routine work from someone's week, those hours "could simply become five more hours of work. AI doesn't decide between those outcomes. The organization does."

What does it cost when nobody decides?

Skills you don't notice losing until you need them.

In the same Dayforce survey, 65% say people are losing the ability to perform some tasks themselves as AI takes them on, 49% say tasks that used to build work experience are increasingly handled by AI, and 51% say AI is reducing early-career opportunities. Dayforce calls this "experience debt."

I wrote about the junior version of this in Who trains the junior when AI does the junior work?. The saved-time question is the same problem from the other side. If the hour AI frees up never goes to learning, the skills the automated task used to teach just leave.

How should a team decide who gets the saved time?

Write it down before you automate, the same way you write down what the agent is allowed to do.

Name the hour before you automate the task0 / 5
12345
  1. 1
    Pick the taskThe clerical one people already want to hand off
  2. 2
    Estimate the hourRoughly how much time a week it really takes today
  3. 3
    Name the ownerWho gets that time: the person, the team, or the backlog. Say which.
  4. 4
    Name the useBurnout relief, learning, customer time, or more output. One, in writing.
  5. 5
    Check in a monthDid the hour go where you said? If not, it went to the backlog.
// Step 3 is the one nobody writes. It's the only one that decides the answer.

It's fine if the honest answer is "more output." Some weeks that's the right call. The problem is never choosing, then being surprised that every hour saved disappeared into the queue.

So who gets the time?

Whoever writes it down first. Usually that's the backlog.

Pick one task you automated this year. Write one line next to it: who that hour belongs to, and what it's for.

Related essays
More essays

Owned, undistributed, and published only here — no newsletter, no funnel, no upsell.

All essays ›
‹ all essays

Let's build
something joyful.

Agents · Engineering · Talks · Podcasts · Interviews · Automations.