Readying for Agentic Commerce

The way products are discovered, evaluated and purchased is starting to shift in a meaningful way. It's not because of a new channel or format, but because of who (or what) is doing the shopping. It was only a matter of time before AI agents made their way into commerce, offering a new kind of convenience for consumers who want faster, more guided decisions.Learn how companies like Unilever and Walmart are preparing and explore: What Agentic Commerce MeansHow AI Agents Differ From Traditional Search Engines and Retail MediaDiscoverability and Product Visibility in an Agent-Driven WorldGovernance and Trust ConsiderationsCommon Pitfalls to AvoidPractical Steps to PrepareWhat Agentic Commerce MeansWhen it comes to agentic commerce, ultimately, you’re getting to a decision flow that simulates our human experience,” says Ed Johnson, a partner at Deloitte Consulting.There's also an important distinction to be made between agentic agents and generative AI chatbots. “There’s a difference between ‘autonomous’ and performing an action,’’ notes Emily Pfeiffer, principal analyst for commerce technology at Forrester. “If I say, ‘Purchase this for me,’ that’s not autonomy; it’s literally following an order. Autonomous is when [the agent] goes out and discovers something for you based on parameters” it is given.With agentic commerce, an agent stays with the user throughout the entire search and purchase journey all in the agent window, Johnson says.Right now, most AI-enabled shopper activity is in the early stages of the shopping journey, according to McKinsey & Co. Sixty-two percent of respondents in its recent survey reported using AI to compare options — such as brands, models, prices and reviews — making it the most common use case. Fifty-five percent relied on AI to learn more about a category or product, including what features to consider. Nearly half used it for discovery and inspiration, such as generating ideas for what to buy.Fewer consumers reported using AI in later stages of the purchase process, such as building or optimizing their basket, checking out, managing repurchases or seeking post-purchase support, according to McKinsey. Click to learn more from from Publicis Sapient's Peter Szczerba. But there’s no question that the way U.S. consumers shop online is evolving quickly. AI platforms are scrambling to help make the shopper experience more visual, informed and seamless.OpenAI, for example, announced in March it is expanding the Agentic Commerce Protocol (ACP) to make product discovery more relevant and up-to-date in ChatGPT.For CPGs and retailers, “the near-term opportunity lies in ensuring their brands appear prominently in AI-assisted research and comparison — across not only owned channels but also the third-party content ecosystems that AI tools frequently draw from,’’ the McKinsey report states. “Over time, as consumers become more comfortable using AI throughout the purchase process, companies may also need to rethink their strategies for consideration, basket building and loyalty in an AI-mediated world.” How AI Agents Differ From Traditional Search Engines and Retail MediaAgentic commerce will soon shake up how consumers shop online. In a traditional online search, a search engine produces results, usually in the form of links to click to visit another website. The models tend to be driven more often by questions people are asking rather than strands of texts they’re searching on, Johnson says. Search engines will search against the exact words someone types in, whereas agents act in a conversational manner. A consumer is “pushed to another place, but not necessarily maintaining the context of the conversation’’ in a traditional online search, he says.In agentic commerce, “the agentic platform is typically getting significantly more context than you would typically get in a search or an ad surfaced in retail media,’’ says Katherine Black, a partner at global management consulting firm Kearney. For example, instead of searching for hydration gels in a search bar and seeing relevant products that are often displayed by ad dollars, “the consumer is instead sharing that they are dehydrated during long runs,” Black says.The agent then asks for even more context, such as whether the user only wants natural solutions, and what type of climate they are in, before recommending not only a specific range of products but also a routine for using them and other relevant tips, she says. “It is a much richer shopping experience for the consumer than the traditional search-and-see ads that come from the highest bidder with some relevancy.” Retailer networks, on the other hand, have sponsored ads geared at particular brands.“Retail media worked because it matched how people shop,’’ observes Julie Averill, former executive vice president and global CIO of Lululemon. “You paid to show up where attention already was. Sponsored results, featured placements — all of it was built around influencing a human who might be persuaded in the moment.”AI agents don’t behave that way; they are not browsing, they are filtering, says Averill, now CEO of advisory firm Gold Thread.“A good agent is trying to narrow options down based on what the user actually asked for,’’ including price, availability, delivery time and preferences, she says. “If a sponsored product fits, it will show up. If it doesn’t, it simply won’t. The budget behind it doesn’t matter.”In a typical retail journey, a shopper searches, scrolls and is influenced by sponsored placements designed to capture attention. But in an agentic environment, that dynamic changes. When Walmart partnered with OpenAI in March to enable shopping through ChatGPT, it signaled a move away from browsing toward guided decision-making. Instead of navigating a retailer’s site and weighing multiple options, the consumer interacts with an agent that filters choices based on context, preferences and constraints, then routes them to a purchase environment. The point of influence shifts upstream, moving away from paid placement and toward the quality, completeness and relevance of the data that informs the recommendation.Further, AI agents present an opportunity for a more balanced power dynamic between retailers and manufacturers, Johnson says. If the manufacturers understand that consumers are letting the agents do the work, and they expose to the LLMs all of their product data and the benefits, features and claims about their products, “the path to purchase becomes inverted,” Johnson says.Consumers will likely not remain loyal to a particular retailer, but rather, the one that carries the product they want. This gives brands an opportunity to have their product discovered by the AI agents, “and not to rely on the on-retailer site search results,’’ he says. “With agentic commerce, consumers are discovering brands independent of the retailer's platform.”Discoverability and Product Visibility in an Agent-Driven WorldProduct visibility changes when agents, rather than humans, evaluate options, but also, when consumers change their mindset. Consumers have to embrace the transition from SEO to what some are referring to as agentic commerce optimization, Johnson says.That means accepting that the AI agents are becoming increasingly competent in making product recommendations and transactions.In many instances, some of that information is held behind the retailer's firewall; retailers haven’t been willing to share that with consumers until they’ve been authenticated — or with manufacturers, he says.It’s not a simple thing for the AI agent to quickly scan a website and understand the price and relay that back in the chat window to the customer, Johnson says. The data needs to be structured in a way that allows agents to learn about a product with confidence, understand the science behind it and make recommendations to the user, he says.Unlike humans who experience the visual aspects of a product and a website, an AI shopping agent is looking at structured data: price, size availability, shipping time, return policy, material transparency and verified ratings, agrees Averill.“It is not moved by a stunning campaign image or a founder story unless that information has been explicitly factored into the user's preference profile. And it is evaluating all of that in real time,’’ she says. “Stale inventory data, a price that updated this morning but has not propagated to your product feed, a size that shows available but has been out of stock for two days — agents catch all of it, and they move on.”Most retailers struggle to maintain consistent product data across their own enterprise, which Averill says, “has always been a messy internal problem.” Echoing Johnson, she says that this convolutes the shopping experience in an agentic commerce world. “Agents cannot buy what they cannot accurately read. If your product data has inconsistent sizing, incomplete material information, poor taxonomy or a lag between what is true and what is published, your products get filtered out before a human ever sees the recommendation.”The brands that invested in data infrastructure and real-time data pipelines before it felt urgent are going to have a head start that will be very hard to close, Averill maintains.Governance and Trust ConsiderationsConsumer trust will play a key role in agentic commerce. As AI agents become part of how people shop, consumers will have more pointed questions, such as who the agents are working for, why it showed them a particular product and, most importantly, whether they can trust it, Averill says.Organizations must be clear on where they will and will not use agents and be transparent with consumers. “People should know when they are interacting with an AI agent and what it is actually doing on their behalf,’’ Averill says. ‘“Filtered under $100 and ships in two days’ is clear. ‘Curated for you’ is not.”If a brand is getting preferential treatment inside that system, that should also be visible, Averill says. “There is a difference between personalization and influence that is not disclosed. Regulators will care about that. Customers will, too.”Agents should also support human judgment, not replace it. “That is not just about ethics. It is about resilience,’’ she says. “Systems will get things wrong. What matters is whether there is a clear path to correct them.’’Averill says there was no clean roadmap when Lululemon started experimenting with AI agents, conversational tools and early forms of agent-based inventory optimization. Some of it worked better than expected, while some taught hard lessons.Common Pitfalls to AvoidThe biggest mistake Black sees a lot of organizations making is treating AI and agentic commerce, in particular, “as an IT or a marketing topic instead of looking at the impact to the entire business model. For many, especially in retail, there are big implications, and no one silo can solve them in their entirety.” That broader impact is already changing how teams work. At Unilever, employees are shifting away from repetitive, data-heavy tasks and toward roles that require stronger judgment, cross-functional thinking and system-level awareness as agents take on more of the operational load, Reema Jain, Unilever CIO, tells CGT.With organizations still in the pilot phase of agentic commerce, it behooves them to consider what all of the implications might be. One thing that is clear to Averill is that “whoever controls the agents now has the power. The retailers who understand that are building their own. The ones who do not are renting someone else's relationship with their customer.” That is the real pilot underway, she says, and it is not a technology test, “it is a strategic bet.”When building agent readiness, brand leaders should not assume this is primarily a retailer challenge, rather than a CPG responsibility, Johnson says. “Large language models draw on both retailer and product-level information to power discovery,’’ he stresses. “This creates a significant opportunity for CPGs to shape how their products are represented — expanding awareness of key features, benefits and differentiators.”Leaders should also not rely solely on branded websites for product information. Even small changes to website structure can disrupt how AI agents interpret content, Johnson notes. “Establishing more stable, structured access points — such as APIs that provide product information, pricing and inventory — can improve consistency and reliability in how agents surface and recommend products.”Practical Steps to Prepare for Agentic CommerceAs tempting as it may be to want to appear ahead of the curve, organizations shouldn’t rush into agentic commerce, Forrester’s Pfeiffer says. They need to start with a strategic plan and framework of what makes the most sense for their business. “Remember that all of this is a tool, not a goal,’’ she says.Brands may be in a hurry, but consumers are not. U.S. and U.K. consumers believe brands will market to consumer agents, but few trust these consumer agents to act on their behalf — even for low-stakes purchases, according to a January 2026 Forrester blog.Pfeiffer also recommends that brands lean heavily on their software vendors. "They’re working so hard right now to develop things.”Averill says brands should not give agentic systems full decision-making power, even if they look capable early. Like Pfeiffer, she says they should not get caught up in the rush to agentic commerce. “The teams that get into trouble are usually the ones that move too quickly past the supervised phase. It works until it doesn’t, and then you are dealing with something you don’t fully understand.”Technology is not the hard part, she adds. It's ensuring that people building the systems feel comfortable questioning what they are doing or raising a concern when something feels off.“At the end of the day, the accountability does not move. The system produces an output. The company owns the outcome," Averill says. “If a customer gets bad information, it does not matter how the model arrived there, it is still your responsibility. That clarity tends to shape how carefully teams design and monitor these systems.”

Readying for Agentic Commerce

Related Topics

Business Topic

Explore the CG Sales & Marketing Tech Summit