In 1987 They Said They Couldn’t See Computers in the Productivity Numbers. They Recently Said the Same About AI.

Have you been investing in artificial intelligence for months — maybe years — and still don’t have a number to show your board? You’re not alone, and that doesn’t necessarily mean your strategy is wrong.

In 1987, Nobel Prize-winning economist Robert Solow wrote a phrase that became famous: “you can see the computer age everywhere but in the productivity statistics.” The paradox was real. Companies had spent more than a decade buying computers, installing systems, training people — and the economy’s aggregate productivity still wasn’t moving. It came to be called the Solow Paradox.

If you read last week’s column, this will sound familiar: it’s exactly what happened with the electric motor between 1880 and 1920. The motor existed, but productivity didn’t show up in the statistics because factories hadn’t yet been redesigned. The same thing happened with computers in the ’80s, and with the internet through much of the ’90s: companies had websites, email, internal systems, and the numbers still reflected nothing extraordinary. The pattern isn’t new. It’s, in fact, predictable.

And here’s the part that gives this column its urgency: a few months ago, economists at some of Wall Street’s most-followed firms started dusting off that same 1987 phrase — this time applied to artificial intelligence. The argument is almost identical to Solow’s: AI is everywhere, in every corporate presentation and every earnings call, except in employment, productivity, or inflation data. Nearly forty years later, the same observation, about a completely different technology.

If your AI ROI still isn’t appearing, the problem is most likely not the technology. The bottleneck is almost always organizational, not technical. Just like the factories that kept the old steam shaft under the new electric motor, most companies installed AI on top of processes, hierarchies, and approval flows designed for a world without it — and then wonder why it doesn’t deliver what was promised.

Before concluding that AI isn’t working in your company, it’s worth asking yourself three questions.

First: is this really a technology problem, or a process problem? If the honest answer is that the workflow would still be slow even if the AI were perfect, the problem was never the tool.

Second: what decisions in your company are still being made exactly the same way as in 2022? If the answer is “almost all of them,” there’s your central shaft that nobody has dismantled yet.

Third, and most uncomfortable: who in your organization would lose power if AI were truly integrated into how decisions are made? That question tends to explain more silent resistance than any technology adoption report can capture.

Your AI ROI probably isn’t behind schedule. It’s exactly where economic history says it should be: in the stage where the motor has already been installed, but the factory hasn’t yet. The question isn’t when the results will appear. It’s when you’re going to start redesigning what surrounds the tool.

Which of those three questions made you most uncomfortable?