At The Spectator, Ross Clark offers a skeptical note:

At some point there has to be an “emperor has no clothes” moment. Why, when we are in the middle of a boom in Artificial Intelligence, which is supposed to transform the way we do nearly everything, is there no tangible increase in productivity across the economy? Surely, if the claims for AI were even nearly true it would by now be showing up in the statistics for Total Factor Productivity (TFP) published by the San Francisco Fed. TPF is a measure of output relative to input from labour and capital; a measure, in other words, of how efficiently workers are using the tools available to them.

In theory, it should be increasing dramatically with AI as workers use it to speed up tasks and get more out of their day. Trouble is, that does not appear to be happening. Instead, TPF has fallen to zero. It is markedly different from the late 1990s and early 2000s, when computing technology was fast evolving, the internet was becoming available for all businesses – and TFP was rising by around 2 percent a year. 

“AI isn’t making us more productive,” October 7, 2026

He thinks he knows what one of the problems might be:

Is the technology of the computer age failing to deliver productivity gains because it is too distracting? Clearly, there are many ways in which AI could help workers improve their output. But there also multiple ways in which it could reduce output. If workers are spending their time playing around with AI, getting it to produce videos of cats on bicycles, or whatever, it is going to damage their concentration and detract from their output. The same problem did not occur with, say, the machinery employed on production lines. A power saw, to take an example, is not much of a source of entertainment.

“Isn’t making us more productive”

Another problem may be this…

Once the boss has dug deep for brand new AI, along with all the training, there’s going to be some sense that we must all at least try to use it, even if it’s not really what we should be doing right now.

Let’s say that there is a fellow downstairs, hollering that he is angry about repeated frustrations with the current customer service chatbot. Chances are, he has a complex problem to resolve. He needs an actual person to read the file, listen to him, and walk him through a dispute resolution process.

He is free to sue, if he likes, but there is a good chance that focused attention to the account can resolve the matter by the close of business and save him the trouble.

He is most unlikely to be placated if, instead of that, we tell him that we have just developed a brand new chatbot for him to go home and try out…

In short, whether AI makes us more productive probably depends in large part of how we decide to use it.