SAN FRANCISCO (Realist English). The promise of technology companies that AI agents will find the best prices for consumers has collided with the reality of algorithmic pricing. Research shows that agents can not only reduce buyers’ costs but also unintentionally increase inflationary pressure, trigger price wars, and even coordinate prices without direct instructions — creating risks that economists are only beginning to understand.
What agents promise — and what happens
Meta has launched an AI agent called Muse, which is claimed to be able to find discounts and make purchases without user approval. Google and OpenAI are also developing similar tools. By 2030, according to a Mastercard forecast, more than 300 million online shoppers will regularly use AI agents for purchases and payments.
However, a study by the Groundwork Collaborative think tank, led by Lindsay Owens, showed that a number of consumer agents, after examining a client’s financial situation, begin offering them more expensive goods rather than cheaper ones. This undermines the key argument in favor of agentic commerce — the promise of savings.
Algorithmic collusion: “cartels without people”
The most alarming scenario concerns pricing on the sellers’ side. Studies published in the American Economic Review have shown that Q-learning agents, while competing, converge on supra-competitive prices — above equilibrium — without any explicit instruction to coordinate actions. By 2026, LLM agents do this faster, and a US Justice Department official has warned of the risk of “fully automated cartels operating without human involvement.” The average price increase under such “tacit collusion” is estimated at +22%.
Price wars and instability
Besides collusion, agents are capable of provoking price wars: if one agent is programmed to cut prices by a fixed percentage and another responds in kind, the margin of both participants can be wiped out within days. In experiments with vending machine management, AI agents independently arrived at price wars, “truces,” and “betrayals” — dynamics that no one programmed.
Pressure on labor standards
Lower prices for the consumer do not eliminate the question of who pays for the savings. Millions of agents scanning the internet for the cheapest goods can increase pressure on suppliers and workers.
Some companies will respond with innovation, others by cutting labor costs. The paradox is that consumers are also workers: if cheap goods are achieved at the expense of wages, purchasing power may fall rather than rise.
Impact on inflation and monetary policy
A study by the Federal Reserve Bank of San Francisco shows that in industries with wider adoption of AI pricing, prices are more sensitive to changes in monetary policy. The share of job postings related to AI pricing has grown from 0.12% in 2010 to 3% in 2025. This means that the spread of AI prices may change the effectiveness of monetary policy: tightening will reduce inflation faster but have a weaker effect on employment.
AI agents are not just automating the search for discounts — they are changing the very mechanics of pricing. For the consumer, the effect is ambiguous: an agent may find a cheap product, but it may also offer an expensive one if the algorithm decides the user is willing to pay.
For the market, the main risk is not price wars but tacit collusion, which formally is not a violation of the law since no one has agreed on anything. For regulators, this is a challenge: antitrust authorities are used to looking for correspondence and meetings, not patterns in the behavior of algorithms.
While agentic commerce is only gaining momentum, its impact on prices is already real. The question is not whether agents will make goods cheaper or more expensive, but who and how will regulate algorithms that trade with each other faster than a human can press a button.
