AI Could Make More Prices Negotiable

A price displayed beside a product simplifies the transaction. The seller offers terms and the buyer decides whether to accept them. That arrangement can remain attractive even when both parties have sophisticated software, because avoiding negotiation saves effort and makes offers easier to compare. AI does not need to displace posted prices to change where bargaining becomes worthwhile.

The relevant distinction is between accepting an announced offer and participating in the formation of the terms. A posted price can change throughout the day without becoming negotiable. A personalised price can reflect information about a customer without giving that customer any additional influence. Bargaining requires a process in which both sides can propose changes and decide whether the resulting exchange suits them.

Shahidi and colleagues examine how agents could change the costs of searching, communicating and contracting in digital markets. Their analysis describes possible market designs and the trade-offs surrounding them, rather than documenting the disappearance of posted pricing. Its relevance is that reducing the effort of arranging a transaction can make different forms of exchange feasible.^[1]

The Competition and Markets Authority's March 2026 discussion similarly considers agents that compare offers, negotiate and switch services within a user's instructions. It presents opportunities alongside questions about reliability, interoperability and incentives. These are prospective capabilities and scenarios, not evidence that autonomous bargaining already improves every consumer's outcome.^[2]

Consider a hypothetical buyer who wants a television but can accept delivery next month and cares more about a longer warranty than rapid dispatch. A seller with inventory arriving later might value that flexibility. An automated exchange could discover terms both prefer to the initial offer. The potential gain comes from matching the parties' requirements more effectively, not from the mere fact that machines exchange counteroffers.

There is an obvious alternative: the seller could publish several packages. The buyer or its agent could choose between immediate delivery, a discounted later date and an optional warranty without bargaining. A well-designed menu might capture most of the benefit more cheaply and transparently. Comparing negotiation only with one inflexible price would therefore exaggerate its advantage.

Bargaining becomes more interesting where relevant combinations are difficult to specify in advance, circumstances change quickly or information has to be elicited before a suitable package can be formed. An unusual mix of quantity, timing and support might justify an exchange that ordinary published options cannot accommodate conveniently. The additional value still has to exceed the costs of negotiating and administering the resulting agreement.

AI could lower part of that cost by reducing the human effort required to discover and evaluate proposals. It would not remove every cost. Systems need usable information, compatible terms and clear limits on what they can commit. A seller receiving many speculative requests also needs a way to distinguish worthwhile negotiations from activity that consumes resources without a realistic prospect of trade.

For buyers, credible alternatives matter more than conversational sophistication. A system able to compare several suppliers and complete a purchase elsewhere can bring a meaningful outside option to the exchange. Repeatedly asking a captive supplier for a discount does not create the same position. The market's competitive structure remains important even when the bargaining itself is automated.

Agency creates a further relationship to examine. The CMA notes that a consumer agent might steer choices towards the provider's commercial interests rather than the user's preferences. A lower displayed price would not necessarily be a better deal if the system sacrificed service, flexibility or suitability that the buyer valued. Understanding how the agent is rewarded is part of understanding the transaction.^[2]

Research on online markets also cautions against treating more elaborate price discovery as an inevitable consequence of digitisation. Einav and colleagues studied eBay's shift from auctions towards posted prices, examining the trade-off between discovery and convenience. Auctions are not bilateral negotiation, but the result is a useful counterexample to the assumption that cheaper digital interaction must make buyers prefer a more complicated process.^[3]

Buyer agents could strengthen posted-price competition instead. A system that searches efficiently and purchases the best suitable advertised package may reward sellers offering clear terms without negotiation. The same technology could thus increase bargaining in one market and improve the attractiveness of published offers in another. Which outcome emerges depends on the problem the transaction presents.

There are also two different sources of apparent gain. If a flexible delivery date reduces the resources needed to fulfil an order, the exchange can create additional joint value. If an agent simply obtains a lower price for the same service at the same underlying cost, it may mainly transfer part of the existing surplus from seller to buyer. Both can matter to the buyer, but they are not the same economic effect.

Nor does more personalised bargaining guarantee that the buyer receives the larger share. A capable seller's system might infer urgency or willingness to pay and defend its terms more effectively. The relevant assessment includes the quality of the match, the resources consumed and the distribution of the gains, rather than treating the number of negotiated discounts as a complete measure of consumer benefit.

The legal framework continues to apply to the transaction. In the UK, the CMA's guidance on the unfair-commercial-practices provisions of the Digital Markets, Competition and Consumers Act 2024 explains requirements for clear pricing, including unavoidable charges. Using an agent does not make an opaque price transparent. That is a limited point about consumer-pricing obligations, not a complete account of how every agent-mediated contract is formed.^[4]

Competition concerns also remain distinct from the efficiency of bargaining. The CMA's discussion of AI and collusion distinguishes established coordination involving agreements from questions about autonomous systems learning behaviour that softens competition. It identifies risks for investigation, not a finding that every pricing algorithm coordinates unlawfully. Sophisticated transaction software does not by itself establish a competitive market.^[5]

A plausible direction is therefore selective expansion rather than the end of the price label. Transactions with several adjustable terms or valuable continuing relationships may offer enough additional value to support bargaining. Standardised purchases made primarily for speed and certainty may continue to favour posted prices or menus. Those are expectations about economic conditions to test, not a universal forecast for particular industries.

The consequential change would be that some offers currently treated as final become starting points for an exchange, without making every offer negotiable. AI can make bargaining available where the human effort once made it unattractive. Whether the transaction should use that capability depends on what it adds beyond a clear price and a sensible set of options.