Three years into the generative AI boom, most companies still work roughly the way they did before ChatGPT existed. Employees use AI to summarize a meeting, draft a first version of an email, or clean up a spreadsheet formula. Developers lean on coding assistants to move faster through routine work. These are real improvements, but they are small ones. The technology behind the largest infrastructure buildout in tech history has, so far, changed remarkably little about how work actually gets done inside most organizations.
That gap between spending and results is now the central question in enterprise tech. Billions of dollars have gone into AI infrastructure, yet a growing body of research suggests the payoff is concentrated in a narrow slice of companies, while everyone else is still figuring out where to start. Understanding why that gap exists, and who is positioned to close it first, says a lot about where business technology is actually headed for the rest of 2026.
A Small Group of Companies Is Pulling Away
A global study from PwC, which interviewed more than 1,200 senior executives across 25 sectors, found that nearly three quarters of AI’s measurable economic value is being captured by just one fifth of companies. The rest are largely stuck running pilots that never scale into anything that shows up on a balance sheet.
What separates the leaders is not how many AI tools they have deployed. It is how they use them. Top performing companies are two to three times more likely to use AI to chase new revenue opportunities rather than just trim costs, and they are nearly three times more likely to let AI make decisions without a human checking the work first. That combination shows up directly in the numbers: leading companies are generating 7.2 times more value from their AI spending than everyone else, with profit margins running about four percentage points higher.
Part of the reason so many companies are stuck in pilot mode comes down to cost, not ambition. As we covered in our look at why AI compute often costs more than the human workers it was supposed to replace, the economics of running AI at scale are messier than the marketing suggests. Cheaper tokens have not translated into cheaper AI, because the more capable systems now being deployed use far more compute per task than a simple chatbot ever did. Companies without a clear plan for managing that spending end up with a large bill and little to show for it, which helps explain why so many pilots quietly stall.
Economists Are Pushing Their Timelines Back, Not Forward
The World Economic Forum’s chief economists have been tracking expectations for AI driven productivity gains for more than a year now, and the mood has shifted from confident to cautious. EY-Parthenon estimates that AI could eventually lift economy-wide labor productivity somewhere between 1.5 and 3 percent over the next decade, with the largest contributions coming from tech, finance, consulting, legal, and accounting work.
But the timeline for when that shows up has been sliding. Compared with expectations from earlier this year, chief economists now expect meaningful productivity gains to take longer across nearly every industry, with the slowest progress in engineering, construction, utilities, and healthcare. Information technology and education are the only sectors where expectations have held steady rather than cooled. Even so, 92 percent of chief economists surveyed still expect AI adoption to keep growing over the next year. The consensus, roughly speaking, is that the gains are real but not automatic, and definitely not as fast as most companies budgeted for.
None of this means the technology is stalling. Google’s aggressive AI push at I/O 2026 made that clear, with the company weaving AI directly into search, its operating systems, and a brand-new laptop category built around always-on AI agents. What it does mean is that owning powerful models is no longer the differentiator. Knowing how to fold them into everyday operations is.
Why Small Businesses May Get There First
Here is the part that surprises people. The next wave of productivity gains may not come from Fortune 500 companies at all. It may come from small and medium sized businesses that move faster simply because they have less to untangle.
Fifty seven percent of small businesses in the United States are now investing in AI tools, up sharply from 36 percent back in 2023, and roughly 30 percent of employees at these companies now use AI on a daily basis. The tasks being automated are unglamorous but valuable: scheduling, invoicing, proposal generation, customer follow ups, and compliance reporting. Businesses that have adopted automation for these kinds of tasks report productivity gains of 20 to 30 percent within the first year, and sales staff in particular say they are saving more than two hours a day that used to go into manual data entry and calendar juggling.
The reason smaller companies can move quickly is structural. A 200 person firm can swap out its invoicing process in a week. A company with a 20 year old enterprise resource planning system and a dozen departments fighting over budget cannot. That is the same dynamic playing out in the jobs conversation, where some analysts argue AI will accelerate hiring in certain roles rather than eliminate them, particularly in the messy, judgment-heavy work that automation still struggles to handle end to end.
Small businesses are not adopting AI without friction, though. Sixty two percent say they do not fully understand what benefits are actually available to them, and 60 percent say they lack the in-house expertise to implement and maintain the tools they do try. The average small business now juggles a median of five separate AI tools, which is efficient in theory but can quickly turn into its own management headache without someone dedicated to keeping it organized.
The Shift From Chatbot to Co-Worker
The common thread across all of this research is that the biggest gains do not come from having a conversation with an AI model. They come from embedding AI directly into the processes a business already runs every day, so it is doing the work rather than just being asked about it.
That is a harder problem than installing a chatbot, and it explains why the companies pulling ahead are the ones treating AI as an operational overhaul rather than a productivity app. For everyone else, the opportunity has not closed. It has just moved from experimenting with AI to actually rebuilding a workflow around it, which is a slower, less flashy project than the last three years of AI headlines might suggest.
If there is one lesson in the data so far, it is this: the tools are already capable enough to deliver serious gains. What is still catching up is everything around them, the processes, the training, and the willingness to let AI actually own a piece of the work instead of just assisting with it.
