0
0
For about three years now, we have all been told the same story. Artificial intelligence is going to change work. It will make companies faster. It will make some jobs disappear. Get ready.
So somebody finally went and asked the people who would know.
Researchers surveyed nearly 6,000 chief executives, finance chiefs and senior finance managers across four countries, and asked a plain question: what has AI actually done for your company? Not what do you hope. Not what does the industry say. What happened.
Nine in ten said nothing had happened at all.
What Nearly 6,000 Executives Actually Reported
The survey ran between November 2025 and January 2026. Teams at the Federal Reserve Bank of Atlanta, the Bank of England, the Deutsche Bundesbank and Macquarie University asked identical questions of firms in the United States, the United Kingdom, Germany and Australia. The results were published by the National Bureau of Economic Research as working paper 34836, under the flat and honest title Firm Data on AI.
Two findings sit at the front of it.
The first is that businesses have adopted this technology. Sixty-nine percent of the firms said they were using AI right now. In the United States it was seventy-eight percent. Even in Australia, the lowest of the four, it was fifty-nine percent. Around three-quarters expect to be using it within three years. The most common use, by a wide margin, was generating text.
The second is that almost none of them could point to a result. Eighty-nine percent reported no impact on productivity over the past three years. More than ninety percent reported no effect on employment. Among the firms that did report a change, the tilt was slightly positive — the researchers put the average at about 0.29 percent, which is close enough to nothing that it takes a decimal point to see it.
Widespread adoption. Almost no measurable outcome. That is the whole finding, and it has been quoted everywhere.
The AI Productivity Gap Nobody Quotes
Here is the part that gets left out.
The same executives who said the last three years produced nothing were then asked about the next three years. They predicted their firms’ labour productivity would rise by 1.4 percent. They predicted output would rise about 0.8 percent. They predicted employment would fall by 0.7 percent.
Read those two answers next to each other. It did nothing. It is about to do a great deal.
That is not a contradiction, exactly. New tools often take years to show up in the numbers. Electricity took decades. The people running these companies may simply be right, and early.
But it is worth naming what that second answer actually is. It is not a measurement. It is a belief about the future, held by people who just told you the past had not delivered. When you see the headline “ninety percent say AI did nothing,” you are reading a report. When you see the forecast, you are reading hope with a decimal point attached.
One and a Half Hours a Week
Buried further in is a number that reframes everything above it.
Executives reported using AI for an average of about 1.5 hours per week. The American workers surveyed alongside them reported about 1.8 hours.
Sit with that. The technology we have described as the largest shift since the internet is, for the average person at work, being touched for slightly less time than a football match takes to play.
Which raises an obvious possibility that very little of the coverage entertains: maybe AI has not transformed productivity because almost nobody is actually using it very much yet. Not because it cannot. Because they are not. Adoption, it turns out, is a word that covers an enormous range — from a company that has rebuilt its operations around it, to a company where somebody has a tab open.
That is a far less dramatic story than either the boosters or the doubters are telling. It is also the one the data most plainly supports.
The Bosses and the Workers See Two Different Futures
Alongside the executives, the researchers surveyed about 3,000 employees in the United States. Same technology. Same country. Same moment.
The workers expected employment to go up by around half a percent. The executives expected it to fall by more than one percent. On productivity, workers expected modest gains; the executives expected roughly twice as much.
So the people who make the decisions expect this to shrink the workforce and lift output. The people inside the workforce expect the opposite on jobs, and are less impressed by the tool generally.
Neither group has the answer. They cannot both be right. What the gap actually measures is not the future — it is who is standing where. If you have already been through a restructuring, you read this tool one way. If you are the one who signs off on the restructuring, you read it another. This is not the first time the same trend has looked completely different depending on your seat; we saw the same split when seven in ten workers said they got no pay raise while half of employers insisted they had given one.
Where the Cost Actually Lands
One more detail, and it is the one that matters most to an actual person.
When those executives predicted employment would fall, they were mostly not describing layoffs. About two-thirds of the drop they expect comes from hiring fewer people.
That changes who pays. A layoff is loud. There is a date, a meeting, a final message to colleagues. Reduced hiring is silent. It is a job that is never posted. An application that gets no reply. A role that quietly gets absorbed rather than filled.
The person carrying that cost never appears in any survey of firms, because they were never inside a firm to be counted. They just keep applying, and keep hearing nothing, and slowly begin to suspect the problem is them.
It is worth saying clearly: it is not. A hiring freeze is a decision made in a room they were never in, about a budget line, in a year that happened to be this one. It is a fact about the room. It was never a verdict on the person.
And there is something strange in how easily we forget that. We built a way of measuring value that counts what a person produces. Then we built tools that produce things. Now we are surprised to find people quietly wondering what they are still for.
Long before anyone measured productivity, a much older idea was already in circulation — that a person’s worth was never a reading somebody took in the first place. It was given, before there was any output to weigh, and it did not move when the output did. You can disagree about where that idea came from and still notice how badly the modern replacement is holding up.
The numbers in this survey measure firms. They were never able to measure you.
What to Do With This If You Are the One Waiting
Nothing here requires money, or a therapist, or spare time you do not have.
- Separate the market from the mirror. When an application goes unanswered, write down the actual reason it might have gone unanswered — a freeze, a budget, an internal candidate. You are not being optimistic. You are being accurate.
- Do not confuse quiet with rejection. Two-thirds of the expected job loss in this survey is roles never opened. Silence is often the sound of a door that was never built, not one closed in your face.
- Spend an hour, not a season, on the tool. If the average executive uses AI for about ninety minutes a week, you are not as far behind as the noise suggests. An hour of honest practice puts you level with the room.
- Keep one thing in your week that produces nothing. A walk, a meal with someone, a conversation with no agenda. It is the simplest available proof that your worth is not an output figure.
If the last point landed harder than you expected, our free What Is My Purpose? assessment is a short, honest walk through the same question — what you are for, when nobody is measuring.
A Question Worth Sitting With
Nine in ten executives said AI has done nothing for them so far — and then predicted it would do a great deal within three years. Which answer do you think is closer to the truth: are we early, or are we wrong about what this tool is for?
Tell us where you land in the comments. We read them.
If You Want to Share This
Nearly 6,000 executives were asked what AI did for their companies over three years. Nine in ten said: nothing. Then they predicted big gains anyway. The strangest number in the whole survey is that they use it about 1.5 hours a week. https://bgodinspired.com/index.php/personal-growth-and-life-skills/ai-productivity-6000-executives/
Everyone quotes “90% of executives say AI did nothing.” Almost nobody quotes the next line: two-thirds of the job losses they expect come from hiring fewer people, not from layoffs. That cost lands on someone nobody is counting. https://bgodinspired.com/index.php/personal-growth-and-life-skills/ai-productivity-6000-executives/
If you have been applying for months and hearing nothing back, this one is worth a read. The freeze is real, it is measured, and it is not about you. https://bgodinspired.com/index.php/personal-growth-and-life-skills/ai-productivity-6000-executives/
Questions People Are Asking
Has AI actually improved productivity at companies?
According to a survey of nearly 6,000 executives across the United States, United Kingdom, Germany and Australia, conducted between November 2025 and January 2026 by research teams at the Federal Reserve Bank of Atlanta, the Bank of England, the Deutsche Bundesbank and Macquarie University, 89 percent reported no impact on labour productivity from AI over the previous three years, and more than 90 percent reported no effect on employment. Among firms that did report a change, the average was a small positive tilt of roughly 0.29 percent. The findings were published as National Bureau of Economic Research working paper 34836.
How many companies are actually using AI?
In that same survey of nearly 6,000 firms, 69 percent said they were currently using AI. Adoption was highest in the United States at 78 percent and lowest in Australia at 59 percent. About three-quarters of firms expected to be using AI within three years. The most common application was generating text with large language models, reported by 41 percent of firms.
How much time do people actually spend using AI at work?
Less than most coverage implies. In the survey described above, executives reported using AI for an average of about 1.5 hours per week, and the roughly 3,000 American workers surveyed alongside them reported about 1.8 hours per week. This suggests that widespread adoption of AI does not yet mean widespread or intensive daily use.
Will AI cause job losses?
The executives in this survey predicted employment at their firms would fall by an average of 0.7 percent over the next three years, with about two-thirds of that reduction coming from hiring fewer people rather than from layoffs. American workers surveyed at the same time expected the opposite — an employment increase of around 0.5 percent. Both figures are predictions rather than measurements, and the two groups disagree sharply.
Why do executives predict future AI gains if they have not seen any yet?
The survey does not answer this directly, so any explanation is interpretation rather than a finding. One plausible reading is that new general-purpose technologies historically take years to show up in productivity statistics. Another is that current usage is still very light — around 1.5 hours per week for executives — so there may not yet be enough use to produce a measurable effect. A third is simply that the forecast reflects expectation rather than evidence. The survey reports both the flat past results and the optimistic forecasts without reconciling them.
Further reading: our look at why 57% of workers are staying in jobs out of fear rather than loyalty covers the other half of this labour market — the people who already have the job, and will not move.
