Who gets the time AI saves?
Synopsis: AI adoption asks employees to reveal information that can affect their own working lives, so what the organisation does with that information determines what it gets told next.
Late in 2025 I was working with a client's "AI Champions" group. These are the people who've put their hands up to lead their organisation on AI adoption. I encourage them to experiment because I believe an attitude of experimentation - "Dora the Explorer" style - is crucial to working with the moving target of AI.
They found a lot of things that worked, and they were getting pretty good at using AI to increase productivity.
Part-way through a conversation, one of them said something like: "If I save an hour, who benefits? Them or me?"
The productivity part of that question is obvious. What interested me is the trust underneath it, because the trust shapes the culture around everything that follows.
What you need people to tell you
AI adoption programmes ask employees for much the same things: experiment, tell us what works and what doesn't, bring us use cases, business cases, report mistakes, help your colleagues learn.
The issue is you're asking people to hand over information the organisation doesn't currently have, but quite a lot of that information is "dangerous" to the person holding it:
- "I can now do this in half the time."
- "I've been using a tool you didn't know about."
- "The approved tool isn't good enough, so people are working around it."
- "This output looked convincing and it was wrong."
- "The target you've set doesn't make sense any more."
- "I don't think we should automate this."
You can have good technology, sensible policies, keen champions and a healthy training budget, and still be running partly blind if people don't feel able to say those things to you.
I was speaking with a friend of mine last year - he knows someone working in translation of technical manuals. The translator found AI helps him work 10X quicker than his colleagues who weren't using AI. But he delivers work 2X quicker to his boss, because he doesn't want to end up simply being given 10 times more work to do for no additional benefit.
Trust is whether people can reveal useful information without expecting it to be used against them.
Before they tell anyone, employees run a quick prediction: what happens to me if I say this out loud?
There are some good answers.
Maybe the team takes on client work it has been turning away.
Maybe the report gets a proper second pass for the first time in years.
Maybe they spend some of the time learning the tools properly.
Maybe they stop finishing the report at the kitchen table on a Thursday evening.
Or maybe, like our translator's concern, the target simply goes up. Three more hours of work arrive by Friday, and somebody upstairs starts wondering how many people the department needs.
Most employees make a guess which version they'll get, predicted from the evidence they have. They remember what happened the last time somebody finished early, admitted a mistake before the client saw it, challenged a target that made no sense, or found a quicker way of doing something.
If the prediction is bad, the sensible move may be to keep quiet, take the free hours and use them however they like.
That's a completely rational response to the incentives in front of them.
This is where culture comes in. Culture determines what people expect will happen after they do what leadership asked them to do.
Strategy tells people to find a better way to work. Culture tells them what will happen if they do.
How organisations teach people to hide gains
Nobody decides what the saved time is for. I wrote about this in 2025: "Where is the ROI".
The savings arrive in ten- or fifteen-minute fragments broken across the day. The time gets eaten by meetings and the next bottleneck, and six months later the team feels a bit less squeezed but nobody can show what changed. The "only" benefit is better working conditions, less stress, and employees potentially enjoying their work more. How terrible!
I think that's a massive win and is a net positive for organisations. The trouble is nobody can see it, so when the accounting department are looking at the cost of AI licences they're wondering why bother.
If saved time is capacity, it only turns into a monetary return when leadership decides where it should go.
If people expect disclosed efficiencies to become a higher target or tighter deadline, they don't stop using AI. They stop disclosing that they're using AI.
They get very good at it, quietly, and keep the time saving and less stress.
If the reward for revealing a productivity gain is more work, don't be surprised when the next gain stays hidden. They stop showing you the real workflow, telling you what it saves or where it fails, teaching the person beside them, or suggesting that the whole process could be redesigned.
You end up with individual AI capability and no organisational learning, and the irony is the organisation created that outcome itself by making the information unsafe to share.
I wrote about employees hiding their AI use back in 2024. The tools have improved a lot since then, but the reasons for hiding haven't changed a whole lot.
This is all back to trust.
Trust runs in both directions
Employee's version: can I safely tell you that my job got easier?
Management version: do you trust your people with time you're paying for but haven't personally allocated?
If one of your people finishes all of today's work by three o'clock because they've found a better way to do it, what do you want them to do next?
They could take on more work, help a colleague, learn something, fix a process that annoys everyone, think about the problem nobody has had time for, or even go home?
There's no single right answer, and it will differ by business and by week.
If the first answer is always "give them more work", then you've already decided what every AI gain in your business is for. Your people will figure that out.
Efficiency rises, the target rises, the workload returns to the maximum, and then it happens again. Under that arrangement AI can increase output without making anyone's working life better. It lets the organisation run people closer to the limit. The accounting department like this version.
But you keep doing that and people acquire a reason to make the next gain less visible.
Often I'm running six or eight internal projects simultaneously - AI is working on all of them and I'm thinking "look what's possible". I'm self-employed and I reap the entire benefit of that efficiency and capability so I'm very happy to work like that.
But somebody who sees how I'm working may be thinking "is that next year's quota for ME?" With a nervous laugh, they say "Don't show that to my boss" (that's an actual quote and they were deadly serious).
I'm not saying you should completely give up on a return on investment in AI, nor that it's not massively beneficial. None of this requires abandoning measurement. Measure the work: quality, turnaround, rework, errors, what customers experience, how much work is happening after hours.
But I would be much more careful about measuring whether every minute of someone's day has been filled.
A team planned at 100% utilisation is brittle. In manufacturing they typically run at 75-85% capacity - if you schedule a line at 100% of theoretical capacity, you leave no room for these activities without missing plans, so you either skip or defer maintenance and have the associated higher failure risk, or constantly run behind schedule expediting, overtime, chaos.
In teams, headroom is what absorbs a colleague being sick, an awkward client problem, or the great idea that becomes a new product line.
So who gets that free hour?
When I work through this with teams, recovered time usually has a few places it could go.
It can go into more output. Businesses need capacity, margin and growth, and sometimes the right result of an AI improvement is more client work, a backlog cleared or a hire you don't need to make. It just doesn't need to always be the default.
It can go into better quality of work: a second pass, deeper analysis or a root cause that finally gets fixed.
It can go into fewer late nights or a free afternoon, merely a less stressful day and increased employee retention.
It can go into learning something new. Again, the only benefit you might see here is merely an upskilled workforce and higher retention rates (learning is a top-three driver of retention).
Leadership has to decide what the gain is for. There's a commercial case for sharing some of it with the people who found it, as well as a fairness one.
Giving people part of what they found is evidence that you can be trusted with the next thing they find, so what you're buying is disclosure.
Disclosure is how one person's clever workflow becomes a team workflow, how an unofficial experiment becomes something you can test and govern, and how you hear about the risks before a client does.
Vicious or virtuous cycle?
The vicious cycle is management captures every productivity gain, so employees become reluctant to share. Management sees less information and responds with more monitoring and surveillance. More monitoring gives employees more reason to be reluctant, and so it spirals down.
The virtuous cycle can run the other way. Employees disclose what works and what fails, they see that disclosure didn't automatically punish them, and disclose the next thing. Leadership can test it, govern it, teach it and improve it, and AI capability stops being something a few individuals have worked out and becomes something the organisation knows how to do.
"But I'm paying for that hour"
You are.
The business bought the tools, pays the salaries, carries the risk and needs a return. Sometimes the right call will be more output or a straight cost reduction. Sometimes you won't replace someone who leaves, and sometimes the technology will contribute to a decision about the size of a team.
If you bill by the hour, this gets harder. An hour saved on client work is an hour you can't invoice, and your people know their utilisation target as well as you do, so they have a second reason to keep a saving to themselves. That's a pricing question as much as a people one, and I expect to see more professional service firms move towards value pricing, away from hourly billing.
"What's the answer?"
My suggestion isn't simply "be nice to your people and hope for the best" (though I do believe in what goes around comes around).
What I suggest is that you make the bargain explicit. If you want more output, say so. If the aim is cost reduction, say so. If you're trying to create headroom, say that too.
If a workforce change is coming, own it as a workforce decision rather than letting it arrive as "AI adoption". And don't leave an AI champion or a middle manager to explain a bargain that only senior leadership has the authority to make.
People won't like every decision, and they don't need to. They can disagree with a bargain and still trust it, as long as it wasn't left for them to figure out from what happened to the last person who found a better way of working.
Before your next AI programme
Before the next round of "find efficiencies" and "bring us use cases", decide what happens when somebody succeeds, and tell people before they do. What happens to the first hour somebody finds?
It doesn't need to be a company-wide formula. You could decide that some goes to the backlog, some goes into better work and some belongs to the team for learning the tools properly.
You need to make sure to stick to the bargain the first time it's tested, because people will watch those first few cases far more closely than they'll read your AI strategy.
The question I'd start with is simple: if someone on your team found another way to save an hour tomorrow, would they tell their manager?
If the answer is no, I wouldn't start with another training programme. I'd start with the bargain you've made with the people you're asking to experiment. And if you need help with this challenge, and with any aspect of AI adoption, let's chat.