John Laverick

Essay

What are we actually doing with the time AI saves?

31 October 2026

I wrote a post recently about a Microsoft Copilot briefing I’d attended, where much of the conversation had been about how many jobs AI might replace. It generated some interesting responses, including a fair bit of disagreement about whether organisations should use AI to reduce costs or to deliver more with the people they already have. Both are perfectly reasonable objectives, but I’ve been thinking about the discussion since, and I wonder whether we’re overlooking something.

We’re spending an enormous amount of time talking about how productive AI can make people. I’m less convinced we’re spending enough time thinking about what happens to that productivity once we’ve created it.

Suppose we introduce AI across an organisation and it saves everyone an hour a day. That’s a pretty significant improvement, particularly if you employ thousands of people. But we’ve still got the same number of people, working the same hours, and unless something else changes, broadly the same operating costs. Perhaps people get through more work, perhaps they have a slightly less pressured day, or perhaps they spend that extra hour doing things they previously didn’t have time for. Equally, the hour might simply disappear into other activities and nobody really notices.

I suspect we’re going to see quite a lot of this over the next few years, with organisations reporting enormous numbers of hours saved through AI but struggling to demonstrate where those savings have actually gone. There’s also a difference between saving somebody time on individual tasks and creating capacity that can genuinely be used elsewhere. Saving ten minutes here and fifteen minutes there doesn’t automatically give you a useful chunk of time to do something different.

Of course, AI will reduce the number of people needed to do certain things, and I don’t think organisations should feel guilty about taking advantage of that where it makes sense. Particularly in the public sector, where money is tight, reducing costs is a perfectly legitimate objective. But I do worry about starting with a target for how many jobs we want to remove, because once that’s the objective, we inevitably start looking at which existing activities we can automate. And that assumes those activities are the right ones to be doing in the first place.

We’ve spent decades designing processes around the technology and resources we’ve had available. We’ve built teams, approval processes, hand-offs and entire operating models to make those processes work. Now we’ve got technology that potentially changes some of those constraints quite dramatically, surely that creates an opportunity to question the process itself rather than just finding a quicker way to get through it?

An example I keep coming back to is reporting. If somebody spends three hours every Friday pulling together a report, and AI reduces that to twenty minutes, that’s great. But I’d be more interested in why we need the report, what decisions it supports and whether there’s a better way of providing that information. Perhaps the people making the decisions could access the information directly, or perhaps we don’t need the report at all. We’ve saved two hours and forty minutes by introducing AI, but we might save the entire three hours by changing how we work, and potentially improve the outcome at the same time.

I was also interested in an example from IKEA, which has been widely discussed in the context of AI and jobs. IKEA introduced an AI chatbot called Billie to deal with routine customer enquiries. Between 2021 and 2023, it handled millions of interactions, resolving around 47% of the enquiries it received, with reported savings of nearly €13 million. That’s impressive in itself, but IKEA also talked about retraining 8,500 customer-service colleagues, helping them move into areas such as remote interior design, sales and more complex customer support.

I find that part more interesting than the chatbot. Rather than looking only at how many people it no longer needed answering routine questions, IKEA was also looking at what else those people could do. There’s an important qualification here, because IKEA reported €1.3 billion in remote sales in FY2022, but that was revenue from its wider remote sales operation. It would be wrong to suggest that AI generated that revenue, or that the retraining programme was responsible for it. We don’t have enough evidence to make that claim. Nevertheless, the approach is interesting because IKEA wasn’t just introducing technology to do the same work more cheaply; it was also considering how the nature of the work could change.

I’m not suggesting that’s always the right answer. Sometimes the best thing an organisation can do with the capacity it releases is reduce its costs. Sometimes it will be to improve service quality, reduce waiting times or deal with increasing demand, and sometimes it might create the opportunity to do something entirely new. I suspect the starting position of an organisation makes quite a difference here.

If you’re already comfortably meeting demand, then releasing capacity might create a fairly obvious opportunity to reduce costs. But if you’re an organisation where demand consistently exceeds what you can deliver, exactly the same productivity improvement might simply mean getting more done, reducing waiting times or improving the quality of what you deliver. You might not save any money at all, but that doesn’t mean you haven’t achieved something worthwhile. In fact, depending on the service, it might be a considerably better outcome.

The challenge is being clear about what you’re trying to achieve and measuring whether you’ve actually achieved it. I see lots of examples of AI programmes reporting time saved, numbers of users, prompts submitted and tasks completed. Those are useful measures when you’re trying to understand adoption, but they don’t tell us much about whether the organisation is better as a result. Are we delivering services more quickly? Have we reduced errors or improved decisions? Have we reduced the overall cost of doing something, or are we able to deal with demand that we couldn’t previously manage?

And this is where I think things become more difficult. Giving somebody a tool that helps them write emails or summarise documents is relatively straightforward. Changing the way an entire service operates is not. It involves challenging established processes, changing responsibilities, dealing with organisational boundaries and sometimes questioning things we’ve been doing for years.

It’s also not something a technology team can deliver on its own. There’s a temptation to treat AI as another technology programme: buy the tools, deploy them, train people, measure adoption and then look for the return on investment. But if we’re serious about changing how organisations operate, the people responsible for running those organisations need to be leading the conversation. Technology can enable an enormous amount, but it can’t decide what an organisation should stop doing, what its priorities should be or how it wants to use the capacity that’s been created.

There’s a workforce issue here too. If we want people to actively identify opportunities to use AI, we need to think carefully about how we involve them in the changes. That doesn’t mean promising that nobody’s job will change, because I don’t think that would be credible. But if every time somebody finds a better way of doing something the immediate conversation is about how many jobs can be removed, we shouldn’t be surprised if people become less enthusiastic about identifying those opportunities.

I think there’s a huge opportunity to reduce the cost of delivering services through AI, and in many cases we absolutely should. I just don’t want us to confuse that with its full potential. If we simply use AI to make today’s organisations operate a bit more cheaply, we’ll undoubtedly achieve some worthwhile things, but I suspect we’ll also miss quite a lot.

If AI really does make your organisation 20% more productive over the next few years, what are you actually planning to do with that additional capacity? And have you started that conversation yet?

Because I suspect the answer will tell us rather more about whether AI is going to transform an organisation than the number of licences it’s bought.

These are my personal views and don’t necessarily reflect those of my employer.