Almost every business now touches AI in some way. That part of the story is settled. What is not settled, and what a stack of 2026 research keeps confirming, is whether any of it is actually paying off.
A survey of nearly 6,000 CEOs, CFOs and senior executives across the US, UK, Germany and Australia found that most firms are actively using AI, yet close to nine in ten reported no measurable improvement in labor productivity over the past three years. That is not a fringe result. It lines up with a separate MIT study of hundreds of corporate AI projects, which found that only about one in twenty ever produced a measurable financial return.
The MIT researchers were specific about why. It was not that the AI models were too weak for the job. It was that most companies plugged a chatbot or writing tool into an unchanged workflow, rather than rebuilding the workflow around what the tool could actually do. A tool bolted onto an old process rarely changes the outcome of that process.
This pattern has a history. In the 1980s, economist Robert Solow pointed out that computers had shown up everywhere in offices and factories except in the national productivity statistics. That gap did not close because computers got better. It closed over the following decade as companies redesigned how orders, inventory and customer service actually worked, rather than just putting a computer on every desk. AI looks to be following the same slow curve, just faster.
The pattern shows up even at small companies with far less bureaucracy to slow them down. A Goldman Sachs survey of small business owners found that most who use AI report real gains in efficiency, and many expect it to grow revenue. Yet only a small slice say they have fully built it into how the business runs day to day, the same gap seen at large firms, just at a smaller scale. The businesses that saw real wins were oddly specific: a tutor using an AI writing tool to handle client notes and invoices, a shop owner using AI to write product listings faster. Narrow, bolted-on fixes to one task, not a company-wide overhaul.
There is also a management side to this that is easy to miss. Gallup's 2026 workplace data found that employee engagement has barely moved even as AI use has climbed, and access to AI tools alone did nothing to change that. What did make a difference: workers whose managers actively supported and explained the AI rollout reported engagement far above those left to figure it out alone. In other words, the tool is not the bottleneck. The manager is.
For any business leader watching AI spending climb, the lesson from this pile of data is uncomfortable but simple. Buying software licenses is not a strategy. The companies pulling ahead are the ones rebuilding specific tasks, ones with a clear right answer that is easy to check, like customer replies or drafting, around what AI can do, and measuring the actual result rather than guessing at time saved. Everyone else is paying for a tool that sits on top of the same old process, and getting the same old result.