90% of CEOs See No AI Productivity Gains. That Is a Choice.

A study published this month by the National Bureau of Economic Research surveyed 6,000 CEOs, CFOs, and executives across the US, UK, Germany, and Australia. The finding that made headlines: nearly 90% of firms said AI has had no impact on employment or productivity over the last three years.

Fortune called it the return of Solow's productivity paradox — the same phenomenon that plagued early IT adoption in the 1980s, when economist Robert Solow observed: “You can see the computer age everywhere but in the productivity statistics.”

I do not find this funny. I find it tragic.

The Gap Between Promise and Reality

Here is what the study also found: two-thirds of executives are using AI. But their usage amounts to about 1.5 hours per week. And 25% of respondents are not using AI in the workplace at all.

Meanwhile, MIT researchers have documented that AI implementation can increase worker performance by nearly 40%. Corporate AI investment hit $250 billion in 2024. And yet — no productivity gains.

This is not an AI problem. This is an implementation problem.

Why 90% of Companies Are Getting Zero ROI

I talk to business owners every week. The pattern is consistent:

1. They bought a platform instead of solving a problem. Enterprise AI platforms are the new ERP — expensive, overpromised, and underdelivered. Companies buy Copilot licenses for 500 employees because Microsoft told them to. Then nobody uses it beyond “summarize this email.”

2. They delegated AI to IT. AI is not an IT project. It is a business transformation. When the CIO owns AI, you get infrastructure. When the COO owns AI, you get automation that moves the needle.

3. They waited for it to be easy. The companies seeing 40% productivity gains did not wait. They started ugly. They built custom solutions for their specific workflows. They learned what works and what does not. Now they have a 2-3 year compounding advantage.

What the 10% Are Doing Differently

The executives seeing real productivity gains share a pattern:

• They start with one high-value process. Not “AI strategy.” One workflow. One bottleneck. One measurable outcome. They solve the problem first, then expand.
• They build agents, not chatbots. A chatbot answers questions. An agent does work. It monitors systems, processes data, takes action. The difference between a toy and a tool.
• They measure ruthlessly. Hours saved. Errors prevented. Revenue recovered. If you cannot tie the AI investment to a business outcome, you are doing it wrong.
• They treat it as a practice, not a project. AI is not something you implement once. Models improve. Use cases evolve. The winners iterate continuously.

I Am Here to Fix This. One Company at a Time.

I am not interested in selling you a platform. I am interested in finding the three places in your operation where AI will actually move the needle — then building it.

That might be order processing. Customer service triage. Inventory forecasting. Quote generation. The specific answer depends on your business. But the approach is the same: start small, prove ROI in 90 days, then expand.

The 90% statistic is not a commentary on AI. It is a commentary on how companies implement technology. The tools are there. The capability is proven. What is missing is someone who understands both the technology and the business problem well enough to connect them.

If you are in the 90%, you do not have to stay there. But the window is narrowing. Every quarter you wait, someone in your industry is learning what works. By the time AI feels “safe” to adopt, the leaders will be two years ahead.

The technology works. The question is whether you will use it — or watch your competitors figure it out first.