AI Saved You an Hour. Where Did It Go?

Maryland just asked its businesses a question most AI surveys skip.

In a report released October 1, the state surveyed 276 leaders of Maryland companies with at least three people. 91% said their organization uses AI in some form. Among the ones who use it regularly, 92% said it has had a positive effect on productivity.

Then the survey asked how much.

70% said the increase was slight. 21% said it was significant.

So nearly everyone is using it, nearly everyone says it helps, and seven in ten say the help is small. I think that is the most useful number a business owner will see this fall, because it points at a problem that has nothing to do with which tool you bought.

The hour is real

Start with the part nobody argues about. The tools do save time.

Workday published a study in January of 3,200 people who use AI at work. 85% said it saves them between one and seven hours a week. Those were employees at large companies, but nothing about the pattern depends on size. A proposal that took an hour takes twenty minutes. A long email thread gets summarized before your coffee cools.

That matches what you have probably seen in your own office. Somebody on your team is faster at something than they were a year ago.

So if the hour is real, and the gain is slight, the hour went somewhere. It goes to two places.

Some of it goes to fixing what the AI wrote

The same Workday study found that nearly 40% of the time AI saves is lost to rework. Correcting errors, rewriting, checking whether the output can be trusted.

This is the part that can be trained away. A person who knows how to hand the tool the right background, a good example and a clear ask gets something usable on the first try. A person who types one line and hopes spends the saved time cleaning up. Same subscription, very different week.

The rest goes back into the pile

This is the bigger leak, and it is harder to see.

Culture Amp, which runs employee surveys for companies, released a benchmark last month covering roughly 112,000 employees. 71% said AI tools help them feel more productive.

Then it looked at workload. Among the heaviest AI users, 72% said their workload was reasonable for their role. Among people who do not use AI at all, it was 69%. Three points apart.

Amy Lavoie, the company's VP of People Science, said she reads that productivity number as "an unanswered question, because workload scores remain steady." The companies that see a real return, she said, will be "the ones that decided, explicitly, what they were going to stop doing with the time it gave back."

Here is what that looks like at ground level. Say you run a twelve-person landscaping company. Your office manager starts using AI to draft the weekly crew schedule and the follow-ups on estimates. It saves her about forty-five minutes a day.

Nobody decided what the forty-five minutes was for. So it went where unassigned time always goes. A few more emails answered. A longer phone call. The filing that could have waited. At the end of the month you ask her if the AI helps and she says yes, a little.

She is telling the truth. She is also the 70%.

Workday saw the same thing from the company side. When it asked what organizations do with the time AI frees up, about a third, 32%, said they simply increase the workload.

Give the hour a job before you save it

A company with five thousand people needs a committee to decide what to stop doing. You need one conversation. That is a real advantage, and most small companies are not using it.

A few things that work.

Name where the time goes first. Before a tool goes to a person or a team, finish one sentence with them: the time this saves goes to blank. Calling back the estimates that went quiet. Walking the job sites on Thursday. Getting out the door at five. Any of those is a fine answer. Leaving it empty is how you end up with slight.

Take something off the list. If AI now handles the first draft of the weekly report, decide what else comes off that person's plate, or whether the report still needs to exist. Time that is not protected gets filled.

Measure where the time went, not the tool. Logins and prompt counts tell you nothing. Count the callbacks made, the quotes that went out the same day, the Fridays somebody left on time.

Go after the rework. Ask your team one question: where do you end up fixing what the AI gave you? Wherever the answers cluster, a better set of instructions or an hour of hands-on training is usually worth it.

Let people keep some of it. If every minute someone saves turns into another task, they learn that being efficient earns them more work, and they stop telling you what they figured out. Giving a piece of the time back is how you keep the ideas coming.

The 21% have the same tools you do

The Maryland number is not a knock on the technology. One in five of those regular users said the gain was significant, and they are not using some secret product. They are using the same assistants and the same AI buttons inside the same software everyone else has.

The report does not say what separates the two groups, so this next part is my read and not its finding. I think it mostly comes down to whether anyone decided what the time was for.

That decision is free. It takes one meeting, and it is available to a company of eight people the same day they think of it.

If your team is using AI and the gains still feel slight, that is something we work through with companies at Nexera Intelligence. Happy to take a look at where the hours are going.