Economists remain divided over how far—and how quickly—artificial intelligence will prove disinflationary. On one side, AI capital expenditure (capex) is driving massive demand for finite resources and skills, pushing prices up. On the other, AI's productivity benefits could raise efficiency and bring down costs over the longer term.

Jeremy Siegel, emeritus professor of finance at the Wharton School of the University of Pennsylvania, has offered a further theory: AI could help address the affordability crisis through agents—software programs capable of carrying out multi-step tasks on a user's behalf—tasked with making consumers' lives cheaper. AI and data centers may not be hugely popular at present, but any assistance in lowering the cost of living is likely to be welcomed.

U.S. consumers currently face inflation of 3.4% as measured by the Consumer Price Index (CPI), according to the latest data from the Bureau of Labor Statistics. That is significantly ahead of the Federal Reserve's 2% target, and the central bank hiked the U.S. base interest rate this month as a result.

Affordability is also top of mind for voters in the run-up to the midterm elections: a July study from Pew Research showed the economy was the most important issue for voters, with 29% saying they wanted to hear plans to address price increases from congressional candidates. A further 15% said affordability and the cost of living specifically were key issues for them.

Siegel suggests AI can help. Referencing the launch of Meta's personal AI agent, Muse, he wrote for WisdomTree that a "fascinating new force" was emerging, one with "a potentially much broader economic development."

Muse has payment capabilities to enact the goals set by users, Meta announced earlier this month. It could, for example, grocery shop for a recipe saved on Instagram, or proceed with the sale of a car once the right price has been agreed upon.

"Companies in banking, telecommunications, insurance, and other industries have long benefited from customer inertia," Siegel noted. "Consumers frequently stay with an inferior rate or service because switching simply is not worth the effort. An AI agent capable of comparison shopping, negotiating, and switching providers changes that equation."

Dotcom comparison

Meta has some way to go, however. The Guardian reported on Sept. 28 that a Muse agent had conducted the sale of an item on Facebook Marketplace in Toronto. The agent completed the sale and arranged for pickup of the item, sharing the seller's home address without permission or approval of the sale.

The emergence of such a service—whether from a hyperscaler, one of the giant cloud-computing operators, or any other market participant—could shift the macroeconomic picture. "I wrote more than 25 years ago that the internet could intensify price competition by making comparison shopping dramatically easier," Siegel added. "The results were mixed because consumers still had to take action themselves. AI agents potentially remove that final friction.

"If they begin negotiating phone bills, moving deposits toward higher-yielding accounts, or routinely finding cheaper alternatives, they could attack what might be called inertial monopolies," he continued. "That could ultimately be an important competitive, and disinflationary, force across the economy."

Consumer demand is already notable. In July, the Organization for Economic Cooperation and Development (OECD) shared in a paper that a third of individuals across its member countries used AI, increasingly to support financial decision-making such as "assistance with choosing and understanding financial products, budgeting, credit management, investing, and retirement planning."

The OECD paper cautioned, however, that increased financial literacy will be needed if consumers are to use the technology effectively and safely. "Consumers need to know how to ask appropriate questions, how to critically assess personal data requests and the responses they receive," the study explained. "Low levels of financial, digital, and AI literacy could further increase the potential for harm associated with the use of these technologies."

For Siegel's thesis to move from commentary to consumers' bills, agents will need to clear two hurdles already on display: the permission and privacy questions raised by the Toronto episode, and the financial, digital, and AI literacy the OECD says consumers need in order to use such tools effectively and safely.

This story was originally featured on Fortune.com.