Nine out of ten companies say AI has had no measurable effect on their productivity or headcount over the past three years. That's the finding from a National Bureau of Economic Research working paper revised in March 2026, which surveyed nearly 6,000 CEOs, CFOs and other senior executives at firms in the US, UK, Germany and Australia. More than 90% reported no effect on employment, and 89% said the same about labour productivity, measured as sales per employee.
Here's the awkward part: 69% of the firms surveyed are actively using AI. It just isn't showing up where boards usually look first. More than two-thirds of executives said they regularly use AI themselves, but the average use was only 1.5 hours a week. A quarter said they don't use it at work at all.
The same executives still expect a very different future. They forecast a 1.4% lift to productivity over the next three years, with output rising 0.8% and employment falling 0.7%. No proof now. Real impact later. The paper, by Ivan Yotzov, Jose Maria Barrero, Nicholas Bloom, Philip Bunn, Steven J. Davis and other co-authors, echoes Robert Solow's 1987 observation that "you can see the computer age everywhere but in the productivity statistics."
Companies are cutting before the proof arrives
Companies aren't waiting for the gains to show up before reshaping payrolls around them. TechCrunch's layoff coverage this year has tracked companies including Axis and Monday.com tying cuts or hiring shifts to AI. The Financial Times reported that several companies have cited AI in job-cut decisions, while economists warned that the evidence for mass white-collar automation is still thin. A layoff memo can say AI. The operating data can say something else.
Salesforce is a good example because the story is more complicated than the slogan. In April, Salesforce said its AI agent had handled 2.6 million customer conversations on its help site with a 63% resolution rate, and that hundreds of support engineers had been redeployed rather than backfilled. Fortune later reported that its CEO said Salesforce had reduced its customer support workforce from 9,000 to 5,000. That is a serious change, but it still doesn't prove every lost role was cleanly replaced by software. It proves management believes fewer people are needed around that work.
Amazon gives you the same caution from another angle. The company confirmed about 16,000 job cuts in January 2026, according to Computerworld, while saying it would keep hiring in strategic areas. Reuters reported in July that Amazon also cut roles inside its artificial general intelligence group. Even the companies spending heavily on AI infrastructure are trimming in some AI teams while hiring in others. That is not a simple automation story. It is corporate restructuring with AI language attached.
The customer still notices
Klarna remains the cautionary example for executives to keep in mind. In its public filing, the Swedish fintech said its AI-powered assistant handled 69% of customer service chats in the twelve months ended June 30, 2025, did the work equivalent of over 700 full-time agents, and delivered about $39 million in 2024 cost savings. Those are real figures, and they are why the story spread so far.
Then came the human part. Bloomberg reported in May 2025 that Klarna's CEO said the company's push for cost savings in customer service had gone too far, and that customers would always have the option of speaking to a real person. TechCrunch later reported his line at SXL London: "Two things can be true at the same time." AI can reduce the cost of routine support, and customers can still get angry when the company treats support as a cost line instead of a relationship.
Forrester has already put a number on that risk. In its 2026 workforce predictions, it expects 50% of AI-attributed layoffs to be quietly reversed, with some jobs returning offshore or at lower wages. That is not a moral claim. It is an operations warning. If you cut now and redesign later, you may just buy the same labour back in a cheaper and messier form.
The skepticism isn't only coming from bosses. A survey of 4,454 chief executives across 95 companies found 56% had seen neither revenue nor cost benefits from AI, while only 12% reported both. A separate survey of 5,000 workers found 40% of non-managers said AI saved them no time each week, while 19% of executives said it saved them over 12 hours. Seventy percent of non-managers said the same.
Why this matters for customer support professionals
If you work in customer support, the gap between executive forecasts and operating data is your job security in both directions. The Salesforce and Klarna examples show that AI can genuinely reduce headcount around routine conversations - but the Klarna reversal and Forrester's prediction of quietly reversed layoffs show that cutting too deep creates problems companies end up paying to fix. The NBER paper's core finding - that 90% of firms see no measurable gain - means most support teams are not about to be replaced by software. What is changing is how companies measure support work, and the skills they expect from the people doing it. If your employer announces AI-driven cuts, the data says you're more likely to be redeployed than replaced. That's a reason to learn how the AI tools in your AI for Customer Support work - not to panic about your role disappearing.
For those managing support teams, the practical takeaway is to track what the AI actually does before restructuring around it. The AI Learning Path for Call Center Supervisors covers exactly this kind of evaluation. Klarna's numbers looked great on paper and still produced angry customers. The companies that get this right will be the ones that treat AI as a tool for reducing repetitive work while keeping humans on the conversations that build loyalty - not the ones that cut first and measure later.
Your membership also unlocks:
