Where Did the Four Hours Go?

Ask a leader how AI is going and you’ll usually get an answer about time. We’re saving about four hours a week per person. It’s the most common AI result I hear, and it’s almost always said with real pride, because it’s true. The hours did get saved.

Then comes the harder question, and I ask it gently, because I’ve had to answer it myself. Where did those four hours go?

That one tends to land in silence.

Time saved is the result everyone reports first. It’s also the one most likely to turn out to be a story rather than a return, and the difference between those two is worth understanding before you put a number in front of your CFO.

A quick word on scope. AI can create several different kinds of return, from better quality and faster decisions to lower risk, more capacity, and revenue you wouldn’t otherwise have won. It also comes with a longer list of costs than most business cases admit, and not only the subscription line. Both of those deserve their own honest accounting, and they’ll get it in future articles.

This one is about the return people reach for first, and the one I’m asked about most: time saved.

Why We All Reach for Time Saved

There’s a good reason time is the default measure. It’s easy to see, easy to count, and easy to explain in a status update. Nobody needs a data team to notice that a task which took ninety minutes now takes twenty.

It also feels like progress, and that matters more than we tend to admit, because early AI adoption runs on belief as much as evidence.

I use the measure too. The trouble starts when saved time is where the measuring stops.

MIT’s Project NANDA studied hundreds of enterprise AI efforts and found that 95% of organizations were getting zero return on their GenAI investments. A lot of the companies inside that 95% can tell you precisely how many hours they’ve saved.

Saved Time Is a Claim, Not Yet a Return

Here’s what I’ve come to think after watching a lot of AI programs up close. Hours don’t bank themselves.

Unless somebody decides in advance what the reclaimed time is for, it disperses. It goes into more email, a longer meeting, a slightly less rushed afternoon. All perfectly pleasant, and none of it visible anywhere a finance team would look.

Time saved is an input, the raw material of a return. What you do with the hours is what shows up in the numbers.

Saved time doesn’t bank itself. It leaks.

The Question That Does the Work: So What?

The most useful tool I’ve found for testing an AI result is a short and slightly rude question. You take your result, and you ask, so what?

Then you take the answer and ask it again. And again, until you land somewhere that matters.

It’s awkward the first few times, partly because it sounds dismissive, and partly because the first answer always seems like enough on its own. Push through that. You’re not being skeptical about the work, you’re being curious about the consequence, the way a good CFO is curious. It helps to ask it of your own projects before somebody else asks it of you.

If you’ve used the Five Whys to chase a problem backward to its cause, this is the same instinct pointed the other way, forward, toward value.

Here’s what it looks like in practice.

Six rungs down, four hours a week finally becomes revenue you kept, which is a number your finance team already tracks.

Most of us stop at rung two or three. “Managers stop rewatching recordings” is a genuine improvement, and it’s also completely invisible to the business. The fourth, fifth and sixth answers are the ones that connect the work to money.

You’ve gone far enough when you reach money. Revenue you won, revenue you kept, or cost you removed. If your ladder runs out before you get there, that’s still a finding worth having. Either the value is real and you haven’t traced it yet, or it isn’t there, and you’ve learned that cheaply.

Most of us stop three rungs too early, then call it a business case.

Two Questions That Keep the Number Honest

Once you’ve traced the value, two more questions stop the time number itself from flattering you.

Where did the hours actually go? This is the redeployment question, and it’s best answered in advance. Saved to do what, exactly? More prospecting calls, faster proposals, an earlier finish for a team that’s been running hot. Any of those can be a legitimate answer, including the last one, as long as somebody chose it. Time that nobody assigns tends to get spent on whatever is loudest. (In my week, that’s usually email.)

What did it cost to check? We’re good at measuring what AI produces and much less good at measuring what it costs to trust it. If the AI drafts an analysis in four minutes and a person then spends ninety minutes verifying it, your real saving is whatever survives the checking, and some weeks that’s nothing at all. Measure net time rather than the headline number. That last stretch of human judgment is where a lot of the quality lives, so I’m not suggesting you skip it, only that you count it. There’s more on that balance in the 20-60-20 playbook.

Making the Number Mean Something

None of this needs a dashboard. It needs four things written down before you start, and about twenty minutes of honesty at the end.

If you can’t say where the hours went, you didn’t save time. You misplaced it.

Your Homework

Take the AI result you’re proudest of, the one with the best time number attached, and run it down the ladder. Keep asking so what until you reach money or run out of answers.

If you reach money, you’ve found your business case, and it’ll be considerably more persuasive than the hours were.

If you run out, you’ve found something almost as useful, which is a project that needs redesigning or retiring before it takes up another quarter.

Either way, write down where those hours have actually been going. That answer on its own tends to change what people do next.

• • •

Getting a leadership team from “we’re saving time” to “here’s what it earned us” is one of my favorite sessions to run. If that sounds like a conversation worth having at your company, drop me a note.