If the reward for saving time is more work, people will stop saving time

In my last role I worked closely with office HR managers across EMEA. In a lot of offices that was one person, sometimes one and a half, covering recruitment, onboarding, leavers, performance reviews and salary reviews, with everything else landing on top. If AI is going to give time back to anyone in a People function, it should be people like them, because so much of their day is the simple stuff: the handbook look-up, the email asking how many days' holiday someone has left.

So say AI takes some of that off their plate and saves them an hour. What happens to it? Unless someone has decided, it goes where spare time always goes in a stretched team, into more of the same. That matters more than it sounds, because of what it teaches people about using AI at all.

The reward problem

I qualified as a Prosci® Certified Change Practitioner this year. The ADKAR model at the centre of the methodology ends with Reinforcement, which is whatever makes a new behaviour worth repeating. The rollouts I saw put their effort into the front end, with licences and a training session, and left Reinforcement to chance.

But Reinforcement happens whether you plan it or not. If someone saves an hour with AI and the hour is immediately filled with more work, they've learned that saving time earns them more work. That's still a reward, just the wrong one, and people respond to it. Some will stop bothering. Others will keep using the tools and stop mentioning it.

I saw what the right kind of reward looks like when Copilot came in at the firm. I was in one of the early rollout waves, and a senior colleague in my team, who didn't see what the fuss was about, went to the hour-long training because that was the condition of getting a licence. Not much of it stuck. What changed her mind was sitting down together. I showed her how I was structuring my prompts, and then we worked on one of our own documents until it produced something she could use. The payoff was in her own work, straight away, and that's what made it worth doing again.

People will hide the time

Anyone who has managed a budget knows what happens to an underspend near year-end. It gets spent, because handing it back risks a smaller budget next year. I've seen it happen, and I don't blame the people doing it, because it's a rational response to how budgets are run.

Saved time works the same way. If people suspect that showing an hour saved means losing it, or losing a colleague, they'll absorb it and say nothing. The organisation ends up with a licence count, a few anecdotes and very little idea of what AI is actually doing for it.

The savings are murkier than they look anyway. Glean's 2026 Work AI Index surveyed 6,000 workers in the UK, US and Australia. They believed AI saved them around 11 hours a week, but they were spending about 6.4 hours a week supervising it and fixing its output, and only 13% said their organisation was performing significantly better because of it. Not every use of AI saves time, so it's worth setting a baseline before rollout and measuring the whole task, checking and fixing included. The biggest gains I've had came from redesigning the process. I automated most of my market research and it went from half a day to around 20 minutes, but only once I'd rebuilt the process around what the tool could do.

What deciding looks like

I've seen an organisation make this kind of decision about time, just not for AI. I created and ran a champions network for inclusion, diversity and equity across EMEA, which grew to more than 70 people covering the vast majority of our offices. Offices setting up inclusion committees were expected to have at least two champions on them, so that at least one person from each office could join our call every six weeks or so and take what they heard back, and I fed what came out of those calls to the senior partner who chaired our global committee. Over time offices started sharing what they were doing and picking up the same projects, which meant we could amplify the work.

What made it last was that the firm decided the time counted. It brought in a contribution policy, so people doing inclusion work were recognised for it, and for the partners who were champions it was tracked in their annual reviews. That's the decision I'd want to see made about AI: time spent helping colleagues use it counts, and time it saves goes somewhere people can see. That decision sits with HR, because HR owns the objectives, reviews and recognition that decide whether saving time gets rewarded at all.

It also helped that not every champion was a believer. One partner joined precisely because they were sceptical. They wanted to be on the inside, to understand what the fuss was about and see what they could learn. They saw the point of the work, but they weren't going to take it on trust. I'd want a few people like that in any AI champions group, because they'll ask the questions others are holding back.

Decide, and say it out loud

So someone has to decide what saved time is for, and then tell people. It could go on more time with the business and the strategic work people rarely get to, or on more time off, or it could mean fewer roles, and any of those can be a legitimate choice. What does the damage is not deciding, or deciding and keeping it to yourself, because people will fill the silence with their own assumptions, and the assumption they reach for is rarely the generous one. Putting the headcount option on the table is uncomfortable for a People function, but it's more honest than leaving people to guess.

None of this works without the basics. For People teams in Europe, the EU AI Act treats AI used in recruitment, promotion and dismissal decisions as high-risk. Those obligations now start in December 2027, but the duty to support AI literacy among staff using these tools has applied since February 2025. So the plan, the policy, the guardrails and data that's safe and secure come first, and they matter more in People than almost anywhere, given what we hold.

After that, adoption comes down to a question every leadership team should be able to answer: what was the time meant for, and did it go there? If the honest answer is more work, don't be surprised when people stop saving it.