Agentic Commerce Is Rewriting The Rules Of Product Search

Welcome to the brave new world of agentic commerce, the next big thing after the web and mobile and marketplace-type platforms. In agentic commerce, agents complete searches for consumers, uncovering more precise product selections with the assistance of artificial intelligence and richer descriptions from shoppers. Retailers can ignore the trend at their own peril, said Scott Eckert, CEO of Mirakl for the Americas, a retail software company.

AI shopping agents are reshaping digital commerce, autonomously influencing consumer purchasing decisions, and redefining online shopping transactions. Call it Commerce 3.0. With agentic commerce, retailers face significant threats, including the loss of direct customer engagement, diminished brand loyalty, and growing reliance on third-party AI platforms. To succeed, retailers must optimize their visibility on third-party AI platforms, develop proprietary AI-driven customer experiences, and build robust operational foundations for AI integration.

“Today we’re very much in the early stages of AI agents being the place customers are starting to go in order to do their e-commerce shopping search,” said Eckert. “In the historical world, people would type a three-word search such as ‘Halloween Table Decorations’ on a Google search or others. In the AI world, we’re seeing people typically typing 11 words or more.”

For example, the Halloween search prompt becomes Halloween-Themed Table Decorations for an Elegant Adult Dinner Party, Eckert said, noting, “You would get a very different answer if you typed that into the search bar. The AI agents are able to understand all that context and your intent from that rich search. That’s what’s changing about how people are thinking about shopping online. Agentic commerce is able to be much more highly personalized because it has all that information from the search request. If it has rich data on products that it’s looking at, it can be very specific about finding products that meet that context-rich search. That’s what’s very different with agentic search.”

While these are early days, adoption is ramping up quickly, Eckert said. “Each of these AI agents operates in a different way,” he added. “You need to optimize uniquely for Open AI [ChatGPT] or Perplexity or Anthropic. All of them have different ways that their agents look at the product information on the retailer’s site.”

Mirakl’s Nexus solution optimizes retailers’ catalogs. It uniquely appeals to each of the AI agents to make sure their products appear in an agentic search, Eckert said. Signs that the transformation is nigh include PayPal launching its Agentic Toolkit and Visa and Mastercard racing to develop payment tools that enable AI agents to make purchases on behalf of customers. Visa introduced Intelligent Commerce and Mastercard debuted Agent Pay.

How fast is agentic commerce catching on? “We went from effectively zero consumer shopping traffic through these AI agents at the beginning of the year, and now it’s somewhere between 2 percent to 5 percent of traffic going from AI agents to retailers’ web sites,” Eckert said. “It’s still small, but zero to a couple of percentage points in less than a year is rapid growth and the expectation is that this will continue to grow and easily get into double digit percentages by the end of next year.”

What exactly is Mirakl Nexus? It’s a catalog transformer, said Eckert. “What it does is it enables a retailer load their product data and the software can actually look at an image of a product, which is a typical component of product information, and pull data out of that image. It can look at a shoe and tell that it’s black and record the data that it’s black. Interpreting data out of images, looking at all the content that has already been provided and making it much richer, is what Mirakl Nexus does. The catalog transformation is enriching the catalog data. Each of those agents is actually done differently. It’s presenting the data to ChatGPT differently from Perplexity and differently from Anthropic. The Mirakl Nexus tool allows customers to do exactly that.”

The implications are clear. Discovery is no longer a matter of showing up in a search result. For merchants and marketplaces, it’s now about whether their product makes it onto the short list that an agent delivers back to the customer, said Eckert, noting that this is nothing less than the rearchitecture of the shopping experience.

Missing metadata, inconsistent language and incomplete images aren’t nuisances anymore, said Eckert, they represent lost business. Retailers lose visibility into customer data because consideration and discovery are not happening on the web site, they’re happening somewhere else, Eckert said.

Agents are now another channel that retailers must embrace and master. And treating them alike won’t work because they’re different and have different requirements. “Some of our customers are beginning to build their own AI agents on their own web sites,” said Eckert, referring to Lowe’s, which has launched something called MyLo, an AI search agent.

“You can put in a rich and complex search such as, Help Me Build My Back Deck and Provide All the Materials I Might Need,” said Eckert. “That agent converses with the customer and helps them find their solution. Many other retailers are thinking of similar ideas.”

Part of the Mirakl Nexus solution is creating the product availability to feed not only external AI agents but also retailers’ own branded AI agents. “It’s much different in the way it said, ‘Let’s have a conversation about yo0ur needs,’” said Eckert. “That’s very different than typing into a search bar. It’s a great example of what many other retailers, particularly large ones like Lowe’s, are considering as a way to take advantage of AI agents themselves to create a better shopping experience. Lowe’s is just a very forward thinking company that got out early.”

The way this is going to impact e-commerce is not entirely clear right now. “What we’re certainly seeing is this change to the early discovery and consideration part of the process with AI agents,” said Eckert. “What’s interesting and not here yet is what happens when AI agents do 100% of the purchase process. Today the agents are doing discovery and delivering a perfectly curated list of products. The agent will go off and search for you and buy it and it will show up at your door. You’ll trust the agent enough to let it autonomously purchase for you. I don’t know when that happens but that is coming down the road. Today it’s just discovery but eventually it will go all the way to making the purchase for you.”