AI is not a productivity tool. It’s a prioritization tool.

You’re using AI. It’s working. You’re drafting faster, brainstorming larger, answering more.

And somehow, you’re just as busy as before.

That’s not a coincidence. I’ve watched this happen across decades as a Chief of Staff. Every efficiency gain was absorbed by the organization into the next set of quarterly expectations – move faster, achieve more, watch the team burnout.

It’s a pattern 160 years old — and it’s playing out on your calendar right now.

The Tool That Ate Its Own Savings

In 1865, economist William Stanley Jevons noticed something counterintuitive: as steam engines got more efficient, Britain didn’t use less coal. It used dramatically more. Efficiency lowered the cost of coal-powered output. Demand expanded until it absorbed — then exceeded — every gain.

Economists call this the Jevons Paradox. You’re living it.

Email promised to eliminate communication friction. It did. Then we sent exponentially more of it. Smartphones put a computer in your pocket to save trips to your desk. Then the desk followed you everywhere. Calendar software promised to organize your time. Then it filled every gap it organized.

Each tool delivered exactly what it promised. The savings were real. And the system absorbed every minute. The paradox doesn’t apply universally — LED lighting and industrial automation eventually plateaued. But for knowledge work, where demand for output is effectively unlimited, the rebound pattern holds.

AI is faster and more capable than any productivity tool that came before it. Which means the rebound will be proportionally larger — unless you do something different this time.

The Data Is Already In

This isn’t a prediction. It’s already happening.

A 2026 study from UC-Berkeley Haas, published in Harvard Business Review, found that AI didn’t reduce work — it intensified it. Employees worked faster, took on broader scope, and extended work into more hours of the day. Often without being asked.

A global Adecco survey of 35,000 workers confirmed the pattern from the other direction. AI saves workers an average of one hour per day. But only 27% used that time for strategic thinking. Only 26% redirected it to creative work. Nearly a quarter reported simply doing more of the same work.

London School of Economics research put a dollar figure on the prize: AI-trained employees save an average of 7.5 hours per week — a full workday — worth roughly £14,000 (~$19,000 USD) per employee per year. That’s the number everyone quotes in the AI adoption pitch.

Nobody talks about where those hours actually went.

Why Wishing Doesn’t Win

When you reclaim 30 minutes, you don’t experience it as free time. You experience it as capacity. And capacity is never neutral. Your organization has a standing offer for every unallocated hour: another meeting request, another escalation, another deliverable that was already waiting.

You don’t decide to donate the time. The default does it for you.

This is why strong intentions aren’t enough. The research is blunt: intentions alone account for only 20–30% of variance in actual behavior. “I’ll use the saved time more strategically” is not a plan. It’s a wish.

AI Bought You Time. Now What?

Here’s the shift that changes everything: AI is not a productivity tool. It’s a prioritization tool.

Every other AI article right now tells you how to do more — more output, more throughput, more scale. That framing is exactly how you end up working the same hours at a faster pace, which is precisely what the Berkeley Haas research observed.

The more useful question isn’t what can I do more of? It’s: what do I actually want my work to look like?

Most people have no ready answer. So the organization answers for them.

The Adecco data hints at what’s possible: the 27% who redirected AI savings toward strategic thinking didn’t stumble into it. They had somewhere to put the time. The difference between them and everyone else isn’t discipline or seniority. It’s decision architecture.

I’ve seen this work firsthand. While in the Office of the CEO at Microsoft, I built an automation tool that cut meeting execution time by 50%. That freed me to focus on what I actually cared about — strategic planning. The intentional redirect didn’t just benefit me: it produced a 1,100-hour reduction in Senior Leader and CEO meeting time, worth millions in company savings. The business outcome followed the personal one.

The Pre-Commitment Question

The intervention is simple. It takes five minutes. Almost nobody does it.

Before you create or adopt any new AI tool, answer one question in writing:

“If this saves me 30 minutes a day, what specifically will I do with that 30 minutes — and is that something I actually want more of?”

Write it down before you start using the tool.

That’s it. That’s the whole thing.

This works because it’s an implementation intention — the psychological mechanism that converts a vague goal into a specific plan. Research published by the American Psychological Association found that people are roughly 3x more likely to follow through when they specify in advance the when, where, and how. Writing down your answer before adoption is what separates redirecting time from donating it.

The question also forces a more honest reckoning. If you can’t answer it — if you genuinely don’t know what you’d do with the saved time — you’re not ready to move forward. You’re just going to expand your scope, accelerate your pace, and call it productivity.

Your Pre-Commitment Framework

Run this before every new AI tool adoption:

Name the time. Estimate concretely how much time this tool will save per week. Not a range — a number. “Roughly 2 hours” is enough.
Answer the question in writing. “If this saves me 2 hours a week, I will specifically use that time for ___.” This answer should be something you want — not something your organization wants from you. Strategic thinking, deeper work, fewer hours, a project you’ve been deferring. Be specific.
Block it before you launch the tool. Put the time on your calendar before you start saving it. A placeholder is fine. The point is to create a destination before capacity disappears into the system.
Review at 30 days. Check whether the saved time went where you planned. If it didn’t, that’s not failure — it’s information. Either the tool didn’t deliver, or the default won. Both are worth knowing.

The Bottom Line

AI will deliver on the efficiency promise. LSE, the Fed, and Adecco all confirm it. The question was never whether AI could save you time.

The question is whether you’ll be the one who decides what happens to it.

Every productivity tool in history gave time back to people who had no plan for it. The organization filled the vacuum. This time, the tools are powerful enough that the rebound could be significant — a full workday per week, quietly recaptured by the system before you notice it’s gone.

You don’t need more AI tools. You need one decision made in advance.

The good news, the hours are coming. Have somewhere for them to go.

Like this article? You might also enjoy:

Written in collaboration with AI

Return On Minute