Hours saved is a common way to describe what AI is worth, but those hours rarely appear in a company's results. Saved time is always used for something. It can become profit when the company pays for fewer hours, lower prices when the saving is passed on to customers, or more output when the same team takes on more work. If nobody decides, it goes into other tasks during the day, and nobody can say what it was worth. In our view that is what usually happens when a business case counts hours and assumes they will turn into money on their own.

What the Danish records show

The clearest evidence we know of comes from Denmark. Researchers ran two large surveys of workers, about 25,000 each time, in eleven occupations where AI can help with much of the daily work, and matched the answers to national records of pay and working hours. Workers who used AI chatbots said the tools saved them about 3% of their time, and most said they spent that time on other tasks. Two years after ChatGPT launched, the records showed that any change in their pay or hours was smaller than 2%.

The researchers think it is too early to see the effect, and that changes in pay and hours may take longer to appear. That is a fair reading. The study also covers chatbots from 2023 and 2024, not today's tools, and it measures workers' pay and hours, not company profits. Still, the pattern is worth planning for: the time was saved, it was spent on something, and the numbers did not move. In our view the same tends to happen inside a company when nobody decides where the saved time goes.

Why boards do not see hours saved

Hours are easy to count and easy to add up, which is why AI business cases often use them. Boards look at other numbers: profit, revenue, cash and growth. An hour saved in a team that already has spare capacity changes none of them, and at a firm that bills clients by the hour, fewer hours can mean less revenue. When the benefit is promised in hours but judged in money, a year later nobody can show what the investment delivered.

Deciding where the saved time goes

That decision belongs to leadership, and it should be made before the tools are introduced. If the goal is more output, someone has to name the extra work the team will take on. Profit means deciding which hours the company will stop paying for. Lower prices means deciding how much of the saving goes to customers. Each of these appears in a number the board already reads, so AI's effect can be checked there against where the number stood before. A saving with no decision behind it cannot be seen in any number.

Exhibit 1
The same saved hour, with and without a decision.
Decided
Goes where leadership sends it. Appears in a number the board reads.
Undecided
Spreads into other tasks. Cannot be traced in any number the board reads.
Tempori view · Danish evidence on where saved time went: NBER Working Paper 33777
Saved time appears in the numbers only where someone decided it should go.

The uncomfortable question

When AI saves your team an hour next month, which number in your next board report will show it?