AI agents will complicate the Fed’s ability to spot economic downturns

Investors, policymakers, and CEOs all understand that artificial intelligence will soon have profound impacts on production, labor, and the economy as a whole. Not only will it affect how work is being done, but it will also affect where changes in our production and levels of inflation are first occurring.

Processing Content

And for these reasons, it will be the Federal Reserve’s job to reconsider how it monitors and responds to these changes. If the Fed wishes to be ahead of the curve with respect to how it reacts to changes in the economy it will need to update its tools for addressing inflation, credit in the banking sector, and labor slack, as well as update its method of data analysis and collection.

There are three main things the Fed can do to preemptively respond to the coming AI economy.

First, in order to see where job growth/output is truly contracting or expanding, the Fed is going to need to collect data on and monitor the price of compute (i.e., the use of data clusters) and energy use in this sector.

Second, to view changes in the price levels, the Fed is going to need to have insight into the expanding shadow banking sector that is going to come with stablecoins and other AI agent-related transactions.

Finally, in response to these new forms of economic changes, Fed policymakers are going to need to think beyond their main interest rate policy tools, and develop new methods to influence the use of AI agents in our economy.

To detail each one of these new challenges facing the Fed, let’s start with how AI will affect labor. As AI has been rolled out in various forms within corporations, CEOs have opined on how they see AI changing the role of the worker within their firms. These visions are all unique but most center around a single idea: that the human worker will spend more of their time managing AIs doing different tasks.

That is to say, within the white-collar job space, the worker of the future may manage one AI that is checking the books, another that is doing graphics work for them, another that is setting meetings, another that is note-taking, etc. The role of the worker will be to give tasks to all of these various AI specialist agents and aggregate their output.

Now, what does this mean for the Federal Reserve? Well, as we move through economic cycles, the first sign of a change in the economy will be in the use of these AI agents. In other words, as an economic recession begins, the human manager will use each AI agent less on a per hour or compute basis.

If the Fed does not have insight into the use of AI agents and waits until they see the human manager losing their job (i.e., the traditional way they measure unemployment), they will be far behind the curve — the human manager’s firing will indicate we are already in a severe recession. In order to identify an economic downturn before it gets severe, the Fed will need to monitor AI compute costs and the energy that is going into these data centers.

In a similar manner, the first signs of inflation will also be moving to where AI agents are interacting/transacting. As we task AI agents with more and more of the orders that need to be made on a personal or corporate basis, it will be in this space that prices first begin to tick up or down.

And, if these AI purchasing agents are using new forms of money like stablecoins or crypto to do their transactions, the Fed is going to need to have insight into price movements in this sector. This area of shadow banking may soon become dominated by retail companies that issue their own coins — and hence will be an area of the economy that is not totally transparent to the Fed. If the Fed is not monitoring this new shadow banking sector for price changes or transactions and watching for the price level to change in the traditional economy, it may be too late to react to inflation movements.

Finally, what does all this mean for the tools the Fed uses to implement policy changes? Traditionally, the Fed uses short term interest rates as its main policy tool to contract or expand the money supply (and influence inflation and the unemployment rate). Yet, we do not know how the price of compute nor the use of AI agents will respond to changes in the short-term rate — it is just entirely untested at this point in time.

The first step in developing new Fed tools to address this new AI world will be figuring out what elements affect prices and labor best in this future economy. Perhaps, it will be the cost of energy that really has a first-order effect on the operations of data clusters, yet even this is untested and will need to be explored.

In all, the Fed will need to adapt new tools and gather more data on what AI agents are doing in this new economy. Changes to inflation and labor will first show up in the output of AI agents and if the Fed is not monitoring this, their policy decisions will be far too late and behind the curve in this new world.