36% of AI Users Hide Their Productivity Gains From the Boss

By Daniel Okafor

36% of workers who say AI makes them more productive admit they hide how much it helps. That figure comes from a survey of 6,000 office workers by Glean, a workplace AI company, across the U.S., U.K. and Australia. The number explains the current AI panic better than any technical explanation. People are not afraid of the tool. They are afraid of what happens after it works.

The mechanics are simple. An employee finishes a report in an hour instead of an afternoon. The work checks out. Now comes the choice. Tell the boss, and become the office AI expert with better assignments and maybe a promotion. Stay quiet, and keep the reputation for being unusually fast. Or a third path: the boss studies the output and wonders whether a team of three could become a team of one.

Silence does not guarantee safety either. A colleague might automate part of the same job first.

What do the surveys actually show?

Glean's survey of 6,000 computer-based workers found some respondents reporting more output and better quality work. Among that group, more than a third concealed the scale of the gain.

A 2024 Microsoft and LinkedIn survey points the same direction. More than half of people using AI at work worried that using it on important tasks would make them look replaceable.

The numbers describe a rational response, not a skills gap. The question troubling workers is not how AI works but what happens to them once it does. No product demo answers that.

Have workers faced this bargain before?

Yes, more than a century ago. Frederick Winslow Taylor described the problem in his 1911 book on scientific management. Some factory workers were paid per item. Find a way to make more, and the employer could cut the rate per piece. Workers then had to produce more just to earn the same pay.

Deliberately slowing down was called "soldiering." Taylor wanted to stop it. Yet he recalled telling workers in his machine shop that, in their position, he would resist producing more too.

Swap the machinery. A manager asks the team to demonstrate their best AI shortcuts. The manager sees a learning opportunity. The employee hears a request to explain how the company could manage with fewer people. A round of applause does not settle that worry.

The author of the piece, who leads a Norwegian research center, describes the trap from the inside. He uses AI to work through budgets and draft meeting minutes. The hours it gives back are exactly what a later budget review can take away. A colleague who uses the same shortcut to get home on time faces the same arithmetic: if the reward for sharing is always another assignment, the relief disappears.

What did Norway just sign?

On October 2, Norway's government and eight employer and union organizations signed an AI agreement running through 2029. It calls for employees to be involved when AI is introduced, using existing rights, and for the gains to benefit workers as well as businesses and society.

There is precedent. In the early 1970s, computer scientist Kristen Nygaard worked with Norway's Iron and Metal Workers' Union on the introduction of computers. That project helped establish participatory design: involving the people who will use a system in decisions about it.

The agreement is a statement of intent. It guarantees no pay rise and no share of the savings. Whether it changes behavior at work remains to be seen.

It also rests on something most American workplaces lack. In 2025, 52.6% of Norwegian employees belonged to a union. The U.S. figure was 10%, and 5.9% in the private sector. A U.S. employer can copy the conversation, not the enforcement.

What would make employees share?

The practical question is straightforward. Do employees get a say in what AI means for their jobs, or are they asked for feedback after the decision has been made?

The piece sketches one model: a six-month trial where the company promises participants will not lose pay or jobs because of the trial. Employees help decide how saved time gets used. Managers agree to discuss results before raising targets. Workers might choose paid training or time to clear a backlog that has kept them working late. The company might want to take on more business. The point is to hold that conversation before the savings are committed elsewhere.

Such a promise needs to be written down, with a way for employees to challenge a broken commitment. It will not bind a company that later decides the saving is a smaller team. It can give people a reason to share what they have learned today.

Taylor's answer was higher pay for workers who met management's targets. Today's employers need an equivalent for the person who finds a faster way to work. Whether the boss ever sees that demonstration depends on what happened to the last person who saved the company time.