To win the next agent purchase, win the person after the first sale
A brand can win an AI-driven purchase before the customer has formed a preference for it. In an agent-mediated purchase, an AI system acts on a shopper’s request and may complete the transaction. The opportunity comes afterward: give the person a reason to name the brand when they next ask an agent to shop. That instruction can help decide who gets the next order.
McKinsey describes a growing core of heavy AI users whose shopping habits are becoming more established, and recommends becoming their default brand. McKinsey also has a commercial interest in businesses acting on that advice, since it advises them on changes in consumer behavior. The mechanism is straightforward: when a shopper names a brand, the agent has an explicit preference to follow.
For marketers, “How do I get chosen?” therefore needs a second question: “Once I’m chosen, can I find the person behind the purchase, and give them a reason to come back?” A direct customer relationship gives the brand a way to shape that future instruction. Advertising can still bring in customers, but an agent-mediated transaction may arrive before the relationship begins.
An agent can remove the encounters that once introduced the brand
To see why the relationship may begin later, consider what happens before payment. For two decades, a familiar marketing model took shoppers through search, an advertisement, a brand’s website, browsing, comparison, and purchase. Each stage let the brand explain its offer and become familiar. By the time the shopper bought, a relationship could already exist.
An agent can compress that sequence into one request: the shopper describes what they want, AI narrows the options, and the system may complete the purchase. The customer spends less effort choosing, while the brand loses chances to introduce itself. So businesses that build preference through advertisements, site visits, and browsing may need to do more of that work after the purchase.
How far this shift goes depends on what the shopper delegates. Someone asking ChatGPT, Claude, or Gemini for product ideas or gift recommendations is using AI to help decide. An agent that places an order also handles the transaction. The difference matters because someone receiving recommendations may still browse a brand’s website, whereas a delegated purchase can bypass that encounter.
Those encounters help people remember a brand, but an agent may start a fresh search with the next request. A shopper may recognize a business after seeing its advertisements and browsing its products; the agent may have no allegiance to the previous seller. Winning a selection therefore secures an order, while the customer’s next instruction can still decide who gets the repeat business.
Without those earlier encounters, the order confirmation may be the brand’s first meaningful communication with its customer. It reaches someone who has already bought but may know little about the business or its service. Post-purchase contact then takes on work previously done by advertising and browsing: explaining the brand and establishing a reason to choose it again.
A project in mind?
Schedule a 30-minute meeting with us.
Senior experts helping you move faster across product, engineering, cloud & AI.
AI shopping is spreading, while holiday peaks raise the stakes for retention
Discovery, influence, and checkout show different levels of adoption
The scale of this change depends first on how shoppers use AI. Optimove’s 2026 research found that 72% of U.S. shoppers use AI tools for shopping ideas or gift recommendations, with nearly half doing so regularly. Optimove sells customer-marketing software, so it benefits when businesses invest in personalization and automated customer contact. Its finding places AI in discovery, before shoppers necessarily reach a channel the brand owns.
McKinsey’s international research covers a different population and a broader range of shopping activity. Its newly published survey of roughly 21,000 consumers across 11 markets found that more than half of consumers worldwide use AI to shop. Optimove’s U.S. discovery measure and McKinsey’s international shopping measure describe different activities, so marketers need to read each on its own terms.
Within McKinsey’s research, shoppers’ assessments help explain AI’s influence on selection. McKinsey found that 81% of AI users believe AI performs well at narrowing their options. AI search was also 3.2 times more likely than traditional search to be ranked as the more influential source in a purchase decision. As shoppers give AI more say over their choices, brands have fewer opportunities to shape those choices through browsing.
That influence is extending into transactions. According to McKinsey, the share of U.S. AI users who had bought directly through an AI tool rose from 10% to 17% over a single year. The change amounts to a 70% relative increase, or seven percentage points. Because the population is U.S. AI users, the figures describe purchasing within that group rather than across all shoppers.
Willingness to delegate suggests room for further adoption within the same broad user base. In the leading categories covered by McKinsey’s research, between 21% and 30% of AI users said they would allow AI to complete a purchase, about 1.5 times as many as had already done so. Willingness describes an intention; completed purchases describe behavior. Teams estimating demand need to keep those measures separate.
Holiday forecasts make that distinction commercially important. Agents including Muse and Alexa for Shopping are expected to place orders for millions of people during the holiday season. That expectation gives retailers a reason to prepare for agent-originated orders, while the survey measures help distinguish customers seeking advice from those willing to delegate payment.
Preparation also needs to account for checkout’s practical limits. Agentic checkout, in which an AI agent completes a purchase, remains early and represents a small share of total sales. OpenAI reportedly reduced its instant checkout plans because of implementation hurdles, illustrating how difficult it can be to carry a purchase through to completion. Existing acquisition channels therefore remain important as businesses prepare for transactions that may bypass some of them.
Holiday volumes increase the value of the first relationship
Even at this early stage of agent purchasing, holiday customer volumes make the post-purchase relationship important. Optimove’s analysis of more than 5.5 million customers across leading ecommerce brands found that new customers increased 593% on Black Friday 2025 compared with an average day. The increase on Cyber Monday was 404%. These peaks bring many first-time customers into a business within a short period, increasing the demand for relevant follow-up.
Shoppers’ willingness to try unfamiliar brands adds to the opportunity. For that holiday season, 47% of shoppers said they intended to try new brands. A first order gives those businesses a chance to establish a preference while the customer is open to a new relationship. An increasing portion of first orders is expected to come through agents, making the introduction after payment more important.
The holiday task also includes customers who already know the business. Optimove found that active returning customers increased 359% on Black Friday, while orders increased 290% in its analysis of returning shoppers. Optimove interprets the gap as a sign that many customers returned without immediately purchasing. That reading directs marketers to consider what returning visitors are trying to do, including on visits that produce no immediate order.
Serving both groups requires more than sending everyone the same follow-up. A new customer may need an introduction to the business, while a returning customer may have a need that differs from their previous purchases. Personalized contact can respond to those differences. The first operational question is what the business can learn about the person behind each transaction or visit.
Use post-purchase contact to understand the person’s current intent
Understanding that person starts by connecting the order with the available customer context. An agent-originated transaction is one signal alongside customer history, preferences, and probable next need. Together, these signals help the brand decide how to begin or continue the relationship. The transaction alone gives an incomplete picture of why the person bought and what would be useful next.
That picture can develop through a sequence of contacts: the order confirmation, the welcome journey that introduces the customer to the business, the initial product recommendation, and the subsequent visit. Each encounter gives the business another chance to learn and respond. Later contact should build on what the business learns.
Purchase history offers a starting point, but current behavior helps explain what the customer needs now. Consider a shopper who has bought running shoes three times. They may be ready to replace a pair, or they may be exploring outdoor jackets for the first time. The same purchase record can sit behind either intention, so the latest action matters when choosing a response.
A recent interest in jackets, for example, gives the business a reason to move beyond a shoe-replacement message. Replenishment would serve one purpose; a recommendation shaped by the new interest would serve another. Personalization becomes useful when that difference changes the next contact. Repeating the customer’s previous purchase category would miss the change in intent.
Making these decisions across customers and channels creates a capacity problem during the holiday peak, described as the busiest week of the year. A team cannot manually examine every latest action, compare it with each person’s history, and choose an individual response at that scale. More customers and more possible intentions increase the workload together. So the business needs a repeatable process that can adjust as behavior changes.
The recommended process begins by gathering real-time customer data and interpreting the latest action alongside historical behavior. AI decisioning, which uses AI to select a response from that context, then chooses the message, offer, channel, and timing. These choices are connected because a useful offer can still arrive through an unsuitable channel or at an inconvenient moment. The system must turn its assessment of current need into an appropriate contact.
This work gives marketing teams a concrete reason to use agents themselves. Agents can help interpret changing needs and coordinate more contacts than people could handle manually. For teams considering this approach, a useful place to focus is the sequence from customer action to response, including the decisions between them. The same process can benefit customers acquired through existing channels and those arriving through shopping agents.
Make commercial information accessible and protect customer records
A customer’s preference can guide an agent only when the agent can find a suitable offer and complete the purchase. McKinsey recommends making product information, pricing, and inventory widely accessible so agents can recommend products and transact. Those details let the system assess whether a business can meet the request. A named preference still needs available inventory and an offer that fits the shopper’s requirements.
Customer records need a different access policy because they describe the relationship behind that preference. McKinsey recommends greater protection for customer information, purchase history, behavioral history, and loyalty data. These records capture what the person has done and how their preferences have developed. Separating them from product, price, and inventory information lets a brand support agent purchasing while retaining greater control over its customer relationships.
Final thoughts
An agent can bring a business a customer who has spent little time choosing the brand. The commercial opportunity is to turn that first transaction into a preference the person expresses next time. A shopper who names the brand gives the agent a clearer instruction, though the offer must still meet their needs.
For business leaders, this means preparing for agent purchasing without neglecting the channels that deliver most sales today. Product, pricing, and inventory information need to be accessible to agents. Post-purchase communication needs to help customers understand the business and respond to their current intent. Customer records require greater protection than the commercial information used to complete a purchase.
The executive question is whether these capabilities work together. Can the business connect an incoming order with the available customer context, choose a useful next contact, and adjust when the person’s needs change? Answering that question will require coordination across marketing, commerce, technology, and data teams.
Agentic checkout remains early, so investment should reflect actual adoption. But improving the relationship after payment has value across acquisition channels. The aim is to give customers a reason to choose the business again, whether they place the next order themselves or ask an agent to do it.
A project in mind?
Schedule a 30-minute meeting with us.
Senior experts helping you move faster across product, engineering, cloud & AI.


