Why AI Has Not Changed the Economy Yet | Callum Williams artwork

Why AI Has Not Changed the Economy Yet | Callum Williams

MTS

October 11, 2026

The Economist's Callum Williams analyzes why generative AI hasn't transformed the broader economy after four years, pointing to historic technology diffusion delays, low business adoption, and a lack of labor market reallocation.

0:00

Callum Williams

Now, you look at the Census Bureau, which does the best, I think, the best survey of AI use in American businesses, it's like 25% of businesses.

After four years, like, it's quite low. I don't think if we'd sat here in 2023 and said, in the far off lands of like late 2026, how many businesses will be using AI? I bet you most people would have said pretty much all of them, and it's now 25%, right?

0:21

SPEAKER_2

All right, we are back with Callum Williams, who is the Senior Economics Writer and San Francisco Bureau Chief at The Economist. We love The Economist or Economist Subscribers.

0:31

SPEAKER_3

Yeah, we love The Economy as well.

0:34

SPEAKER_2

We love The Economist.

0:36

SPEAKER_3

We love Economists.

0:37

Callum Williams

Not many people do these days.

0:38

SPEAKER_2

I love The Economy, I love Economists. I saw a really good tweet a while ago that was like, trust the science MFs when the science is economics.

0:47

Callum Williams

Definitely shouldn't do that.

0:48

SPEAKER_2

Yeah, I trust the science when the science is economics too. So yeah, why is AI not completely changed the economy yet?

0:58

Callum Williams

Well, the history of new technologies is that they take a very long time to diffuse across the economy. This was true for electricity, it was true for the internet, it was true for tractors in American agriculture, which took 50 years really to be used in a serious way in farms. So if you have a even fairly rudimentary grasp of history, you would be entirely unsurprised that four years into the AI, generative AI world, not much has changed.

1:31

SPEAKER_3

Yeah. Well, what about... Sort of.

1:34

SPEAKER_2

Diffusion is very rapid with AI compared to other technologies, for example.

1:38

SPEAKER_3

What about the internet, which did have pretty rapid diffusion throughout society, all things considered?

1:45

Callum Williams

Well, the internet is more capital intensive for a consumer than AI, because you have to buy a modem and hook it up and pay for a subscription, whereas with AI, you can just log in with your browser. So people thought that AI would spread much more quickly than it has.

So I think the way to think about it, so a few things. One is the internet actually didn't spread that quickly when it came to businesses. It did spread fairly quickly when it came to households. But even now, if you look at the statistics, like a very weirdly large share of businesses across Europe and the US, like don't even have a website, for example, just to use a really, really facile example. When it comes to AI, I think what people came to realize was that consumer adoption of AI, minimal consumer adoption was very rapid.

ChatGPT was the fastest thing to a billion users and so forth. But I think when you're thinking about the effect on the economy, you have to think about use by businesses and also not just like occasional chatbot use but the entire reworking of an entire company's workflows to incorporate this fundamentally new technology into how they do stuff. And companies now are just often quite big and bureaucratic and badly managed and it takes time for that to change.

3:05

SPEAKER_2

I don't think it's true that it is that big of a burden to adopt AI in a company, at least to some extent. It's like any employee of a company can like with a browser and a laptop, like get ChachiBT, get Claude Cowork or ChachiBT work today to take over at least some of their tasks, to help them with some of what they do for work. I don't think it's that difficult.

3:27

Callum Williams

Well, so there's a few things there. One is being here, in Berkeley or San Francisco, we massively overestimate how much knowledge there is of AI across the economy, even across the US as a whole.

Most people do not use AI basically at all. They just don't really know what it is. So you can count those people out immediately. Then you have people who use it from time to time. But again, most ChatGPT queries are like, give me a recipe for tomato sauce. It's not how do I convert on customer outreach 15% faster than before. It's very, very basic stuff. And then you get to the people who maybe like people here are more like power users.

Without the kind of like workflow redesign that I mentioned, I think it's quite hard to actually delegate enough tasks to a bot such that you can see any kind of meaningful productivity uplift. And that is indeed what you see from the best, the highest quality surveys on AI and productivity from a team at Stanford and the Bank of England and a few other central banks, which basically find that they go out and they ask many, many, many CFOs what has been the impact of AI on your bottom line. And essentially, it's basically balanced to zero. Now, to be clear, I'm not saying it will always be zero. What I am saying is that there are big, big structural things standing in the way of rapid change.

Read the other 10 minutes of this episode here.

$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.

Not what you expected? Full refund within 14 days.

Thousands of transcripts fetched by people building searchable podcast archives

Fetch transcripts with code

The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000795190264