The Anthropic Economic Index

A couple of weeks ago, Anthropic released something they're calling the Anthropic Economic Index. This interesting and somewhat ambitious project has been designed to try and measure the economic impact of AI adoption across industries; in real time.

Typically, when we're looking at economic impact, we have a collection of what is known as lagging indicators. This means that you can take a view on growth, for example; the last quarter.

It does depend on the metrics in use but typically you might look at a month or a quarter. Sometimes a half year (6 months) or even a full year.

Whatever timeline you review you are looking back on what's happened. What Anthropic is attempting is to track related metrics…

… IN REAL TIME.

Now we do need to take some of this with a pinch of salt. It’s clearly an experiment based on their own data segmentation.  It’s also based entirely on users of Anthropic’s product ecosystem, which will certainly add some bias.

So we'll see how much value there really is here over time… 

… but I thought it was interesting enough to be worth a mention. 

Three preliminary takeaways.

So a few the key takeaways that I thought were worth exploring.  

AI is deflationary. What do we mean by that?

It seems that industries that embrace AI within their workflows see their cost base decline. AI drives intelligent automation and as such improves efficiency. 

It's improving the speed at which their day to day workflows are completed and improving the accuracy and efficacy of work done by employees…

… people have to do less of the work because it's going through some form of AI augmentation. 

The second thing to notice is that the impact is specifically on white collar productivity. 

So knowledge workers.

The use of LLMs and, the ecosystem of tools that are built upon them, are speeding up knowledge worker tasks. 

We're talking about things such as customer service, research activity, data analysis, legal work, finance work, software development, and report writing. 

These things are being accelerated. 

Individuals that are engaged in this activity are become more productive; their productivity is going up and up and up.

Thirdly, something to consider is that this impact is completely uneven. 

While AI is enhancing the tier of highly skilled knowledge work, low skilled automation is in the very early stages. Therefore, only industries that are focused on white collar, knowledge worker based, activities are seeing real impact from the current wave of artificial intelligence and the tools that are available.

Pace of change

Obviously, the normal economic indicators often struggle to capture what's happening with AI in real time. This current wave of technological disruption is happening so fast that looking back on what happened last quarter is of little value.

The world will have radically changed and moved on since then. With new challenges and new opportunities. We are in the phase of development where something that looks like science fiction; wowing us on a Monday, is old hat and out of date by the Friday.

If Anthropic’s approach proves successful then it could be of great value to policymakers, and business leaders. Enabling more informed decisions about AI adoption and the trends impacting organisations now. 

As I said at the beginning, take it with a pinch of salt… 

This data is all from Anthropic’s ecosystem.

The coming wave

Also, obviously the bias is very much in what's happening very, very recently. 

It wants to be a real time tool. Therefore all the data is from the current generation of tooling …  individuals and organisations that are making use of automations for activities such as reporting, writing emails or cleaning up text more efficiently. 

… the Generative AI wave that we've seen to date is where all the benefits are.

We are now moving into the next wave. 

… AGENTS …

Agents will change the landscape again with AI that is autonomous, free thinking, and actively solving problems.

This emerging tool set will be less deflationary as the cost, at least initially, will be significantly more expensive than the cheap / free tooling of the GenAI phase. 

The impact however, of Agentic tooling, will be significant.