A new survey indicates that AI has not yet altered the way most companies operate. This is despite massive investment and claims of revolutionary change, as most business leaders report little change as yet.
The new survey is from the National Bureau of Economic Research (NBER) in Massachusetts. It surveyed nearly 6,000 top executives in the United States, the United Kingdom, Germany, and Australia. The executives included CFOs, CEOs, and managers of companies of all sizes.
Most companies are already using some type of AI. Nearly 69 percent are actively using AI, and three-fourths expect to use it in the next three years. The most common uses of AI are text generation using large language models, image generation, and machine learning for data analysis.
However, the findings indicate a disparity between adoption and outcomes.
The AI Disconnect
Over 90 percent of the executives believed that AI had not affected employment in their companies over the last three years. Close to 89 percent also indicated that there were no changes in productivity, which was measured by sales per employee.
The executives, however, believe that changes will happen in the near future. NBER believes that AI will influence the employment of 1.75 million workers in the four countries by 2028. The executives also believe that productivity will increase by 1.4 percent over the next three years. This will reverse the decline in productivity growth that has been experienced in advanced economies over the years. However, the increase is still small.
Employees, however, have a different perspective. They believe that AI will lead to an increase in jobs compared to what the executives believe. They also believe that productivity will increase by a smaller percentage.
Other research paints a similar picture. A survey by consulting firm PwC found that more than half of business leaders saw neither higher revenue nor lower costs from AI projects. Another study from Deloitte showed that while 74 percent of firms hope AI will drive revenue growth, only 20 percent have achieved that goal.
Massive Investment vs. Elusive Productivity
Even large pilot programs show mixed results. A trial of Microsoft’s workplace assistant found that some tasks became faster while others slowed down. The overall effect on productivity was close to zero.
Jared Spataro, who leads AI at Work efforts at Microsoft, has said it remains hard to measure return on investment because knowledge work does not always link to clear financial metrics.
At the same time, strong predictions continue. Microsoft AI chief Mustafa Suleyman has argued that most computer-based tasks could face automation within about 18 months, including accounting, legal work, marketing, and project management. Such claims stand in sharp contrast with survey data that show slow gains so far.
Hardware and infrastructure investments also move ahead. Companies such as Lenovo report strong optimism among enterprise buyers, many of whom expect positive returns despite limited proof today. Large industrial groups like Adani Group continue long-term plans for massive AI data center projects.
AI’s Transition from Hype to Incremental Integration
Customer service offers one area where leaders expect clearer change. Research from Gartner shows strong pressure on managers to deploy AI tools. Many firms plan to automate routine support tasks while shifting human agents into roles that require judgment and empathy. Companies expect staff to learn new skills rather than disappear from the workforce.
The pattern across studies is consistent. Businesses adopt AI at a steady pace, but measurable gains remain small. Companies experiment, test workflows, and search for clear value. Many leaders believe benefits will arrive later, once tools improve and workers learn how to use them well.
For now, AI looks less like a sudden revolution and more like a slow upgrade to existing systems. The technology shows promise, but the economic payoff remains uncertain. Firms continue to invest, hoping that today’s groundwork will produce tomorrow’s results.
