For generations of shoppers, the price tag has represented something reassuringly simple: this is what an item costs. Artificial intelligence could turn it into an opening offer.
A new Mastercard report predicts that more than 300 million online shoppers could routinely use AI agents to shop and pay by 2030. In some retail categories, those agents could also negotiate.
Go deeper with GlobalData
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
Rather than shoppers asking for a discount, a consumer’s AI could make an offer and a retailer’s AI could counter it – potentially completing the negotiation without either the shopper or a member of staff becoming involved.
That could challenge one of modern retail’s most familiar conventions: the fixed price.
When machines negotiate with machines
Agentic commerce describes a model in which AI does more than recommend a product. An agent can search, compare and transact on a consumer’s behalf, working within instructions and limits set by that individual.
Mastercard’s report,A Short History of the Future of Shopping and Payments, identifies the US, UK, Netherlands, China and South Korea among the markets where AI shopping agents could see significant adoption.
If negotiation becomes part of the process, the dynamics of retail could change.
Fixed pricing helped solve a practical problem for large retailers: shop assistants could not negotiate individually with thousands of customers. AI removes much of that constraint because software can potentially conduct huge numbers of negotiations simultaneously.
A customer’s agent could therefore encounter a £100, $100 or €100 price and make a counteroffer. The retailer’s own AI could respond immediately.
The result would be a new form of machine-to-machine commerce in which negotiation becomes part of the transaction rather than a time-consuming conversation between buyer and seller.
AI negotiation is already being tested
The prospect may sound futuristic, but automated negotiation has already moved beyond the laboratory – albeit on the procurement rather than consumer side of retail.
Mastercard points to Walmart’s use of AI for supplier negotiations. According to figures cited in the report, the system concluded 68% of deals without human involvement, delivered an average gain of around 3%, and 75% of suppliers said they preferred negotiating with the bot.
This is not the same as an AI agent bargaining over a consumer’s checkout price, but it demonstrates that automated commercial negotiation is already technically feasible.
The report also cites Stanford negotiation benchmarks suggesting that the capabilities of the negotiating agent itself can affect the outcome, with weaker buying agents paying roughly 2% more.
That raises an important question for retailers: could the quality of their negotiating AI become a source of competitive advantage?
Today’s ecommerce competition is heavily focused on acquiring customers, optimising conversion and personalising recommendations. In an agentic marketplace, retailers may also need systems capable of deciding when to hold firm, when to add value and which combination of benefits makes commercial sense.
Consumers, meanwhile, could have agents making equally sophisticated calculations on their behalf.
The offer becomes negotiable
For retailers, the obvious concern is margin. If every customer arrives with an automated negotiator programmed to secure the best possible deal, does retail become a race to the lowest price?
Mastercard’s report suggests it does not have to.
Rather than simply reducing the price, a retailer’s AI could construct a different proposition around a customer’s priorities. The headline price might remain unchanged while the offer includes a longer warranty, different service level, loyalty benefit or product bundle.
The report calls this “value calibration”, distinguishing it from simply charging different customers different base prices for the same product. Its prediction is that by 2030, the shelf price could become an opening offer in selected categories, with AI allowing negotiation to take place at scale.
An imagined pharmacy of 2030 illustrates the idea. A customer’s agent seeks a lower price for a blood pressure monitor.
Instead of granting the discount, the pharmacy’s AI counters with the monitor at list price, a free flu vaccination and a loyalty credit on another purchase. The customer’s agent assesses the offer against its instructions and accepts.
The important change is not simply that prices become negotiable. The offer becomes negotiable.
That distinction could be crucial for retailers. Automated negotiation would not necessarily mean lower margins. Instead, businesses could set the boundaries within which their AI operates – defining minimum margins, which products are negotiable and what benefits can be offered instead of a discount.
A consumer’s agent might arrive with its own rules: a maximum price, preferred delivery date, minimum warranty or loyalty preferences.
The transaction starts to look less like choosing a price from a menu and more like two sets of commercial rules interacting.
Retailers will have to sell to algorithms
The implications extend well beyond price.
Retailers have spent decades designing digital commerce for people. Product photography, advertising, promotional banners, brand stories and carefully designed product pages are all intended, at least partly, to capture human attention.
AI agents do not shop in quite the same way.
Mastercard’s report envisages agents checking specifications, availability, warranties and delivery terms directly. They could also assess whether claims about areas such as supply chains, returns and labour practices stand up to scrutiny.
The shopper may therefore not always be the one examining a retailer’s carefully designed product page. Their AI agent may be interrogating the information behind it.
For retailers, the implication is that machine-readable product information could become considerably more important. Specifications, availability, delivery terms, warranties and verifiable claims may increasingly need to be intelligible not only to shoppers but also to the agents acting for them.
Verified claims could become particularly significant. If an AI agent is selecting between competing products, unsupported statements about sustainability, provenance or quality may carry less weight than information that can be independently verified.
Ecommerce could consequently evolve from designing primarily for human discovery towards designing for both human shoppers and machine decision-makers.
Who sets the rules?
Shopping agents could give consumers more bargaining power.
An automated buyer could potentially compare more products, check more information and negotiate more persistently than a person ever could. It could operate without the fatigue and time constraints that limit human shopping.
But retailers can deploy agents too.
Katja Forbes, the futurist who wrote the retail chapter of Mastercard’s report, describes the setting of rules for these machine interactions as “commercial sovereignty”.
Retailers will need to determine what customer agents can ask for, what their own systems are authorised to offer and where the boundaries of a deal lie. In Forbes’s vision, increasingly machine-driven transactions would remain subject to authority and rules established by people and businesses.
That makes the return of haggling very different from its historical counterpart.
Negotiation once required time, confidence and often considerable patience. If AI makes it effectively instantaneous, bargaining could happen on transactions where no human would previously have considered it worthwhile.
Trust will set the limits
None of this is likely to happen at scale unless consumers are willing to delegate meaningful purchasing decisions to software.
Mastercard stresses that trust will be central to agentic commerce. Consumers may be comfortable using AI to find products, compare prices and receive recommendations, but allowing an agent to spend money introduces a much higher threshold.
Agents will need clearly defined permissions and secure payment mechanisms, while consumers will need control over preferences, spending limits and the circumstances in which human approval is required.
Mastercard argues that securing an agentic transaction is not only about protecting payment credentials but also establishing that an agent is acting within the intentions and boundaries set by its user.
Retailers face their own trust questions. They will need confidence that a purchasing agent genuinely represents the consumer, that its authority is clear and that automated negotiations remain within agreed parameters.
The transition is also unlikely to happen evenly around the world. Adoption of shopping agents is expected to vary between markets, influenced by factors including existing levels of confidence in ecommerce and digital services.
For international retailers, the conventional fixed-price model is therefore unlikely to disappear overnight – or everywhere at once.
The return of negotiation
Agentic commerce would not necessarily spell the end of fixed pricing. But it could change what a displayed price represents in parts of retail: not always the final cost, but a starting point for an automated negotiation over price, service, loyalty benefits and other forms of value.
Retail has spent generations turning a price into a statement: this is what you pay.
Agentic commerce could turn it back into the beginning of a conversation.
Except this time, the customer may never hear the conversation at all.
