The Productivity Paradox

So we have 1000 PhD’s in our pocket now, they say.

We have AI that can pass MCATs and LSATs, they say.  

And yet 95% of AI projects have failed so far.

Companies are pouring billions of dollars into investment in AI yet only 6% have seen a ROI. Are we still stuck in the tough of disillusionment? It’s been 4 years since the launch of ChatGPT that started this turbulent AI wave and well, have things taken off?

Let me take you back to the year 1987, before Tik Tok, before Facebook, before the internet. The personal computer revolution was in full swing and companies were spending billions on this amazing new thing called information technology.

Everyone agreed we were entering a new age where workers would become superhumanly efficient. But there was a problem. A brilliant economist named Robert Solow, who would win the Nobel Prize that same year, looked at all the economic data.

He looked at GDP.

He looked at output per worker and he saw nothing.

The massive investment in computers had produced zero statistically significant improvement in productivity.

In fact, productivity growth had actually slowed down.

Solow famously wrote, "You can see the computer age everywhere but in the productivity statistics.”

This became known as the productivity paradox: the baffling economic mystery of why investing in revolutionary technology often fails to make us richer or faster for a very long time. It challenges Silicon Valley's core belief that technology automatically equals progress.

So why does this happen? If you give a writer a computer, they write faster than on a typewriter. If you give an accountant a spreadsheet, they calculate faster. Obviously. But personal efficiency is not the same as systemic productivity.

The main explanation for this paradox comes from something called general purpose technologies, or GPTs. A GPT is a technology so powerful it changes everything: steam power, electricity, the computer. Is it a coincidence that there is also a product called ChatGPT? Yes. Yes it is.

But history shows us that these powerful technologies don't boost productivity right away. In fact, when they are first introduced, they often lower it. Why? Because you have to stop working to learn the new tool. You have to rewire the building, rewrite the manual, and even change the laws. This creates what economists call a J-curve.

You will see more about this in my book called How to do more with less, Future-Proofing yourself in an AI driven economy.

The J-Curve

First, there is the investment phase. You buy the technology, costs go up, and productivity actually dips. Then comes the adjustment phase. You are figuring out how to use the new tool, rewiring your processes around it, and productivity stays flat. Finally, you hit the explosion phase. Society reorganizes itself around the technology and productivity skyrockets.

When Robert Solow looked at the data in 1987, he was deep inside that adjustment phase. He was judging the computer before it had finished changing the world.

The Electric Motor Mistake

Let’s go even further back, to another revolutionary invention – the electric motor, of the 1890s.

The economic historian Paul David noticed something fascinating. The electric motor was invented in the 1880s and by 1900 factories were buying them. But for 40 years until 1920, American manufacturing productivity didn't budge. Electricity did nothing for the economy. The reason came down to factory design.

In the 19th century, factories were powered by massive single steam engines. Because of this, you had to arrange all your machines based on access to power, not logic, not workflow. A huge steel drive shaft ran down the center of the ceiling, and leather belts connected every machine to it. The entire factory had to be built vertically, stacked across multiple stories, just to stay close to the power source. It was dark, dangerous, and incredibly rigid.

When electricity arrived, factory owners made a critical mistake. They simply ripped out the giant steam engine and replaced it with one giant electric motor. They kept the shafts. They kept the belts. They kept the terrible layout. The result? The factory was not faster. It was just quieter. They had electrified the old way of doing things.

It was not until the 1920s that a new generation of managers had a breakthrough. They realized: with small electric motors, we can put a motor on each individual machine. If we do that, we do not need the drive shaft. And if we do not need the shaft, we can arrange machines in the order of the actual workflow. They started building single-story factories with skylights. They invented the assembly line.

But here is the critical detail. Only after they completely redesigned the entire factory floor, a process that took over 30 years, did productivity finally explode. The technology itself was never enough. They needed organizational innovation.

In 1987, companies were doing the exact same thing with computers. Just using them to type old-fashioned memos a little faster instead of reinventing how they worked.

The New Dynamo

So where does that leave us today? We are facing a brand new wave: artificial intelligence. Generative AI is the new steam engine, the new dynamo. And right now, we are hearing the exact same promises. This will double human productivity.

But we are also watching the exact same J-curve play out. Companies are spending billions on AI chips, but they have not figured out how to use them yet. They are jamming AI into old workflows, just like that electric motor bolted into the steam-powered factory.

Lawyers are using AI to write memos faster, but they have not reinvented the legal system. Coders are writing code faster, but they have not reinvented software architecture.

If history and Robert Solow's paradox are any guide, we should not expect a massive productivity boom this year or the next. We should expect a messy, confusing period of adjustment. We will see costs go up. We will see confusion. And we will likely see that initial productivity dip.

The real boom will only come when we stop using new tools to do old things and start doing entirely new things that were previously impossible.

--

What is the progress of your firm when it comes to AI adoption? Are you “removing the steam engine and putting in the electric motor”? Or really rebuilding the factory? If my message aligns with your company – please let me know!

Hello! I’m Sharon Gai, author of How to Do More with Less: Future-Proofing in an AI-Driven World a keynote speaker on AI and its effects on workers and the future of work.