Agent-Based Commerce: Why Price Perception Matters is no longer manufactured in bulk
November 11, 2026
An AI agent does not simply look at a listed price: it compares the actual price—which includes applicable promotions, shipping, availability, and unit price—for the products the customer requests.
Competitiveness is now measured rather than perceived. The accuracy and consistency of pricing data across channels have become the brand's new calling card.
Your price image has been built on the store shelves, based on a few carefully selected products. AI agents that compare and make purchases on behalf of your customers don't shop on the store shelves.
This article is intended for retail pricing, sales, and marketing departments. It explains what a shopper actually compares, why your KPIs are no longer sufficient, and how price perception changes when it is measured rather than merely perceived.
Agent-Based Trading: Definition and How Prices Are Affected
Agent-based shopping refers to purchases in which an artificial intelligence agent searches for, compares, and sometimes buys items on the customer's behalf, based on a request made in everyday language.
Customers no longer browse the aisles or product pages. They delegate the task: “Find me this product at the best price, delivered by Friday.” The agent searches through the offers, sorts them, and presents only a few.
We're talking here about agents who buy, not agents who set prices. That second topic, from the retailer's perspective, falls under the category of agent-based pricing.
For a pricing team, the implication can be summed up in one sentence:price perception is no longer formed solely in a customer’s mind. It is also calculated in a machine’s memory, query after query.
European consumers would entrust an AI agent with finding the best price across retailers—the task they are most willing to delegate. Only 8% would entrust it with payment.
(Deloitte, *The Human and the Agent*, 13,500 consumers, 15 countries, May 2026)
It is no longer a niche application. In the same study, price comparison ranked first among the AI capabilities considered most important for the next two years, cited by 48% of respondents.
Perceived price image, measured price image: the agent perceives nothing
A retailer's price image has always been shaped in part by what's on the shelves: a few well-placed loss leaders, signage, a flyer, and the overall impression left by the checkout experience.
Human customers remember very few prices. They make generalizations based on a handful of reference points they are familiar with, and that is precisely what makes their perception manageable.
An AI agent does not generalize based on a hunch. It compares figures for the specific product being asked about, at the exact moment the question is asked. It has no memory of the atmosphere and no attachment to a particular brand.
We are thus shifting from perceived competitiveness to measured competitiveness. The factors that influenced perception—such as the presentation or repetition of a pricing message—carry less weight when faced with a comparison tool that analyzes only data.
| Criterion | Customer in the aisle | AI Buyer Agent |
|---|---|---|
| What he takes away from it | An overview, a few well-known awards | Statistics, query by query |
| Products Compared | Especially the KVI stores selected by the retailer | The ones the customer requests |
| Special Offers | As seen in stores and in the flyer | Counted only if they are legible |
| Shipping and Inventory | Observed on site | Included in the actual price |
| Pricing Error | Often corrected by the client | Reproduced as is |
| Brand Loyalty | Strong, rooted in habits | None, unless otherwise specified by the client |
Tools for measuring brand image already distinguish between the price index and consumer perception. Agent-based commerce adds a third metric, which is unlike either of the two.
What an AI agent really compares: the actual price
A salesperson doesn't compare a list price. They compare what the customer will actually pay to obtain the product under the requested conditions.
In its guide to agent-based commerce, Stripe describes an agent who searches for shoes within a budget that can be delivered by a certain date, then checks availability, delivery times, and return policies before placing an order.
The actual price therefore takes several factors into account: the listed price, the promotion that actually applies, shipping costs, product availability, and—for products sold by weight or volume—the unit price.
Conditional promotions: the blind spot in comparisons
A discount reserved for cardholders, a “buy 2, get the 3rd free” promotion, or a jackpot prize are not considered prizes as defined by a machine. They are rules, subject to certain conditions.
If these terms and conditions are not published in a clear and consistent manner, the agent compares the base price. The promotional effort exists and is funded, but it is not factored into the comparison.
The opposite risk also exists: a promotion that is poorly described, applied by the sales associate but denied at the time of purchase. In terms of price perception, this is worse than no promotion at all.
Unit price is once again a factor in comparing products
A field that compares two different units converts them to a common unit. Google Merchant Center also provides a dedicated attribute for this purpose, which is used to display the price per kilogram or per liter when comparing products of different sizes.
An incorrectly labeled large-size item, or a price per kilogram that doesn't match the displayed price, skews the comparison. A customer in the aisle can correct this on their own; a store clerk cannot.
The products being compared are no longer the ones the retailer selected
A price-image strategy is based on a choice: the KVI—that selection of products whose prices customers know and in which the retailer invests.
This selection process assumes that the retailer knows which products it will be judged on. When dealing with a sales associate, however, that is no longer the case. It is the customer’s request that determines which product is compared, and it may happen to be a product from the back of the shelf.
A retailer can therefore be very competitive on its key performance indicators (KPIs) yet appear expensive on the products that agents actually inquire about. No one will notice this discrepancy in an index based on the typical shopping basket.
But you still need to know which competing product each SKU is being compared to. Without reliable product matching, it’s impossible to know whether your price is high for a given search query or whether you’re comparing two different products.
Price data consistency is becoming the showcase
If the agent sees only data, the quality of that data becomes the equivalent of the product on the shelf. An incorrect, outdated, or contradictory price is no longer just a minor detail; it is the first impression made.
Three inconsistencies stand out in particular: a different price depending on the channel for no apparent reason; a promotion advertised in one place but missing from another; and a unit price that does not match the listed price.
These price discrepancies are nothing new. We have long argued that price differences between channels should be a matter of choice, not something we are forced to accept. Agent-based commerce makes this requirement visible from the outside.
Companies know this. The Deloitte report, based on an Adobe study, notes that only 44% of organizations consider the quality and accessibility of their data to be sufficient for AI.
Of the French consumers surveyed, have already used an AI tool during their shopping journey to search for, compare, or choose a product.
(Deloitte, *The Human and the Agent*, France panel, May 2026)
It's the same as with an internal pricing engine: the quality of the data determines the limits of what the AI can do. In this case, it also determines what other companies' AI systems will see about you.
The real risk: a race to the bottom based solely on the products being compared
Deloitte warns its readers: if AI-driven discovery is won solely on price, it becomes a race to the bottom. The firm urges retailers to give agents other reasons to choose them.
It’s a familiar temptation: to quickly align with the lowest bidder across the board. It’s costly, because it treats every product under consideration as a key performance indicator (KPI), without taking into account its profitability or its ability to drive sales of other products.
We believe the opposite is true. When dealing with a buyer who is weighing their options, the answer isn’t to lower the price further; it’s to determine at what point a price reduction actually changes their decision.A product’sprice elasticity doesn’t disappear just because a buyer compares it.
That's the difference between reacting to signals and making decisions. We explored this in our article on market signals and profitable decision-making.
The challenge for pricing teams can be summed up as follows: shifting from a price image that is manufactured to one that is demonstrated— product by product, channel by channel—using accurate data.
Would you like to know how an AI agent views your prices?
Our experts work with you to analyze your actual competitiveness— product by product , and channel by channel—including for products you aren't currently monitoring.
Let's discuss your pricing strategyFrequently Asked Questions About Agent-Based Commerce and Price Imaging
Agent-based commerce, sometimes called a-commerce, refers to purchases in which an artificial intelligence agent searches for, compares, and can purchase a product on behalf of the customer.
The customer describes their request in everyday language, specifying a budget, a deadline, or a preference.
The agent reviews the available offers, sorts them, and presents only a few. Depending on the level of autonomy granted, the agent either presents the customer with a selection or places the order directly.
Agent-based commerce focuses on the buyer side: AI agents that compare and make purchases on behalf of consumers or businesses.
Agent-based pricing involves the seller's side: AI agents that propose or set a retailer's prices within limits defined by its teams.
The two scenarios come together: a retailer whose prices are set by rules or by AI has those prices compared by purchasing agents. The two scenarios require different safeguards.
An AI agent compares the actual price rather than the list price.
It takes into account the listed price, the promotion that actually applies, shipping costs and delivery times, availability, and—for products sold by weight or volume—the unit price.
It compares the data to the specific product requested by the customer at the time of the request. Any missing, delayed, or inconsistent data between two channels is accepted as-is, without the corrections a human customer would make on their own.
Only if these benefits are presented in a clear and consistent manner. A discount reserved for cardholders, a prize, or a cumulative reward is a rule with conditions—not a price. If the rule cannot be applied, the agent compares the base price.
The retailer then offers a promotion that isn't included in the comparison. Conversely, a poorly described promotion—one that the sales representative counts toward the total but is then denied at the time of purchase—undermines trust more than the absence of a promotion.
It doesn't eliminate them, but it does limit their impact. A KVI policy assumes that the retailer knows which products its customers use to evaluate it.
When speaking with a sales associate, it is the customer's request that determines which product is compared, including a shelf-stock item.
A retailer can therefore remain very competitive on its key performance indicators (KPIs) while appearing expensive for the products actually surveyed. This discrepancy does not show up in a price index based on the typical shopping basket.
Yes, for a significant portion of them. According to the Deloitte study “The Human and the Agent,” conducted in May 2026 among 13,500 consumers in 15 European countries, 49% of French respondents have already used an AI tool during their shopping experience.
France remains at the bottom of the European rankings, far behind Spain at 67 percent. Use is primarily limited to searching and comparing; delegating payments remains rare, at 8 percent of European respondents.
There is a risk if the comparison is based solely on price.
Deloitte warns that, in this case, AI-driven discovery becomes a race for margins, and urges retailers to give customers other reasons to choose them: service, warranties, quality, and availability.
Pricing everything at the lowest possible level is tantamount to treating every product in question as a loss leader. The profitable approach, rather, is to determine where a price reduction actually influences the purchasing decision—and where it merely erodes profit margins.
Given that its price data is now read by machines.
The priority is accuracy and consistency: a single product must have a consistent price, promotion, and unit price across all channels where it is listed, and any discrepancy must be approved.
Next comes insight into a brand’s actual competitiveness—including factors beyond the KPIs—which requires knowing which competing product each SKU is compared to. Booper assists retailers with this analysis.
Dynamic pricing is permitted in France. Personalized pricing based on automated decisions is allowed, but customers must be notified of it. Algorithm-based price-fixing remains prohibited, even in the absence of contact between competitors.
The strongest safeguard for a retailer: being able to say, for each price, which rule generated it, what data it was based on, and who approved it.
An LLM can match differently worded descriptions and suggest groups of products from the same line. It can also be confidently wrong.
Booper's AI-driven chaining system combines AI, programmed controls that correct its errors, and human validation. No group is created without the user's consent.
An LLM integrated into pricing software processes sensitive data: prices, margins, and volumes. Where the model runs determines who has access to it.
Self-hosted AI keeps the data with the vendor or the client and keeps costs fixed. It does not eliminate the need for checks on each response or for human validation.