90% of CEOs Say AI Had No Impact on Their Business. Here's Why They're Right.

A Fortune article published today reports on a striking finding from the National Bureau of Economic Research: after surveying 6,000 CEOs, CFOs, and executives across the U.S., U.K., Germany, and Australia, nearly 90% of firms reported zero impact from AI on employment or productivity over the past three years.

Companies have poured over $250 billion into AI. The models keep getting better. And yet, by the executives' own admission, almost nothing has changed.

If you're running a business and haven't seen results from AI, this study should make you feel two things: validated, and concerned. Validated because you're not alone. Concerned because the problem isn't AI. It's how you're using it.

The Numbers Behind the Headline

The NBER study doesn't just say "no impact." It explains why. And the details are more revealing than the headline:

• Executives who use AI average 1.5 hours per week. That's 18 minutes a day. You can't transform a business in 18 minutes a day. You can barely check your email in 18 minutes a day.
• 25% of firms report zero AI utilization. Not low usage. Zero. A quarter of companies have invested in AI and aren't using it at all.
• Tool sprawl is actively hurting. A Boston Consulting Group study of 1,488 workers found that productivity increases with 1 to 3 AI tools, then declines when workers juggle four or more. They call it "AI brain fry." More tools doesn't mean more impact. It means more confusion.
• Saved time disappears. Stanford researchers found that workers who do get efficiency gains from AI redirect the saved time to leisure, not productive work. The gains evaporate because no one redesigned the job to capture them.

Read that list again. Every single item is an implementation problem, not a technology problem.

The Solow Paradox, Round Two

Economists are comparing this moment to the "Solow Paradox" of 1987, when Nobel laureate Robert Solow observed: "You can see the computer age everywhere but in the productivity statistics."

He was right. From 1973 to the mid-1990s, U.S. productivity growth dropped from 2.9% to 1.1% per year, even as businesses spent billions on computers. It took nearly two decades for the productivity gains to materialize. When they finally did, between 1995 and 2005, productivity growth surged 1.5% as the internet, networking, and enterprise software matured alongside the hardware.

Apollo's chief economist Torsten Slok put the current moment bluntly: "AI is everywhere except in employment, productivity, or inflation data."

The parallel is real, but it's also misleading if you stop there. The IT boom didn't happen because companies waited long enough. It happened because complementary innovations arrived: the internet connected the computers, ERP systems standardized the workflows, and organizations finally redesigned their processes around the technology instead of layering it on top of existing ones.

That's the part most companies are skipping with AI.

The Counterpoint No One's Talking About

Here's what the Fortune article buries deep in the piece: Q4 2025 showed a 2.7% U.S. productivity jump. Stanford's Erik Brynjolfsson notes that GDP is tracking 3.7% with declining job gains. That's the exact economic signature of a productivity surge beginning.

In other words, the "no impact" narrative might already be outdated. The macro data is starting to move. But it's moving for companies that invested in the workflows and training, not just the tools.

The gap between "bought AI" and "implemented AI" is where the entire story lives.

Why 90% See No Impact (And What the Other 10% Did Differently)

When you look at the data, the 90% who saw no impact share a pattern:

1. They bought tools without redesigning workflows. AI was layered on top of existing processes. Employees got access to ChatGPT or Copilot and were told to "use it." No one mapped which tasks were candidates for AI and which weren't. No one changed how the work actually gets done.
2. They spread too thin. Instead of going deep on one or two use cases, they deployed multiple tools across multiple teams. The BCG data is clear: more than three tools and productivity actually drops. Focus beats breadth.
3. They didn't train their people. The ManpowerGroup 2026 survey found that AI usage among workers increased 13% last year, but confidence in AI plummeted 18%. People are using it more and trusting it less. That's what happens when you hand someone a tool without teaching them how it works.
4. They didn't capture the gains. When AI makes a task 30% faster, what happens to the saved time? If the answer is "nothing specific," the gain is gone. The Stanford data proves it. Efficiency without redesign equals leisure, not productivity.

The 10% who saw results did the opposite: they identified specific, repeatable tasks, deployed AI against those tasks, trained their teams, and restructured workflows to capture the time savings.

What This Means for Your Business

If you've been skeptical about AI's impact, this study proves your instincts were partially right. Most companies aren't getting results. But the reason isn't that AI doesn't work. It's that the way most companies adopt it doesn't work.

Three things to take away:

1. Stop adding tools. Start redesigning tasks. Before you buy another AI product, pick one workflow your team does every day and redesign it around AI. Map the steps. Identify where AI fits. Change the process. That single workflow will teach you more than five unused subscriptions.
2. Depth over breadth. One AI tool, well-implemented, beats four tools deployed loosely. If your team is using more than three AI products, audit which ones are actually driving value and cut the rest. The BCG data says you'll literally perform better.
3. Capture the time. If AI saves your team two hours a week, decide in advance where those hours go. Assign them to a project. Redirect them to customer outreach. Make the gain tangible and trackable, or it will evaporate into longer lunches.

The Solow Paradox resolved itself in the 1990s because organizations finally learned how to use computers properly. The AI productivity paradox will resolve itself too. The question is whether your business learns it now, or scrambles to catch up in five years.