If You Don't Decide Where AI's Saved Time Goes, You Get More Stuff

AI is giving your people hours back, and if you don’t decide where those hours go, you just get more stuff created: more dashboards, more analyses, more documents and presentations, a lot of it made because it was easy to make and with no business need behind it. If AI saves one of your people 8 hours a week, that’s 20% of a 40-hour week, and you’re still paying the same salary for it, right? So the saving doesn’t really show up as cost savings, and most companies haven’t decided what the time is for instead.

Judgment used to come free with the cost of making things

When I was building monthly marketing reports for clients, each one took about a week. Because it took that long, I had to be very intentional before I started about what the point of the report was, what answers it would give and what we were trying to solve with it. Now you can build the same report in an afternoon, and at that price it starts to seem like a no-brainer to just make it, whether it has value or not. Before AI, judgment was essentially free, because making things was expensive.

I’m watching this play out as our clients get good at AI. Across our clients we’re seeing 8 to 12 hours per employee per week saved, which we measure task by task, timing the recurring work a team picks out before AI and again after, and in a lot of areas that’s doubling capacity. That’s a real win. But some of what fills the new capacity exists only because it could be made quickly. It gets passed to someone else, and eventually it ends up on the island of misfit artifacts.

The outside numbers say the same thing. Only 7% of organizations give employees any guidelines on how to use the time AI saves them, according to a 2025 Gartner survey of HR leaders, and in a Danish study of 25,000 workers, a quarter of the people saving time spent more of it on the same tasks the tool had just sped up.

Start from the metric, then find what’s stopping it

More output isn’t necessarily the problem. It’s only a problem when nobody has a strategy for where to point the extra capacity, so start with the business: what are the goals and the KPIs, and what’s actually stopping that number from improving? It’s usually one of three things, and each one wants the hours spent differently.

If it’s capacity, the team knows what works and there’s demand for more of it, they just don’t have the time or the people. A member services team with a backlog of member inquiries and a consistently high resolution rate should spend the hours on more of the same, because more volume improves the outcome.

If it’s quality, the work is getting done but the results aren’t showing up, so more of the same won’t help. For a team that could ship features quickly but was releasing too many bugs, the hours went to testing and code review, and the goal was reliability and a lower cost of fixing problems later.

If it’s untapped opportunity, there’s something valuable the team could be doing and never had the bandwidth for. For one client, that meant having AI build detailed financial models rooted in a business case.

Only the first of those three calls for more of the same output.

I hold my own hours to the same test. A blog post used to take me 6 to 8 hours from scratch, and with AI as a thought partner I’ve cut that in half, so do I write two? I could, and sometimes that’s the right choice, but it isn’t the right choice by default, because something else might be worth more than a second post. My hours went to client work, to building reports that track how our marketing performs, and partly to more content. The reports earned their place because we were launching things blindly without a way to track and monitor them, so they answered a question we actually had. That’s the whole test for me: is there a business need behind it?

Employees make the call, and leaders supply the questions

The obvious pushback is that this is micromanagement, and that the people closest to the work know better than leadership what to do with the time. I agree with most of that. Most decisions about how to use the time AI gives back will and should be made by individual employees and their managers. Leaders set the KPIs, and then three questions go into regular 1:1s and team meetings, asked about whatever the freed-up time is about to go into:

1. Who benefits from this work?
2. What should it change?
3. In the next month, how will I know whether it works?

A month later, revisit the answers against the metric the team is trying to move. It starts at the top, with leaders asking the questions of their own work, and then employees bring them to their own 1:1s. Employees owning the questions matters, because one-on-ones rate highest when the direct report sets the agenda, and a check that feels like the company auditing your hours is how you end up with people hiding their AI use because they expect any gain to turn into more work. The goal is to help people think in the context of the business metrics leaders are trying to drive, so they develop better judgment about where they spend their time.

Sometimes the honest answer to “who benefits” is the employee. If retention or burnout is the number you’re trying to move, giving some of the time back is a legitimate destination, as long as somebody chose it.

Pick the metric, find what’s stopping it, and point the hours there. Skip that and the time goes where it always goes, to whatever is easiest to make.