By: Garett Sloane, AdAge
As AI agents evolve from answering questions to shopping, brands face a new kind of consumer: autonomous software making decisions on someone else’s behalf. To stay visible in this new ecosystem, companies must optimize not for humans, but for the AI agents acting in their place.
Brands were just getting used to generative AI answer engines, and must now consider agents that consumers will use to research and eventually purchase products without ever leaving AI platforms. To accommodate the various agents, brands need to ease their path across the internet, pointing the agents to product catalogs and shopping carts as efficiently as possible, industry experts said.
Brands that don’t assist the assistants could be pushed down in the consideration set the next time an agent comes knocking. For instance, if one brand seamlessly interacts with an agent, that agent could be more inclined to visit the same brand for other consumers, said James Cadwallader, co-founder and CEO of AI marketing technology startup Profound.
“In a future where agents are more successful using certain platforms,” Cadwallader said, “we may start to bias our decisions towards those platforms, creating runaway train scenarios.”
The agentic conversation is clearly high stakes, as brands that cater to agents could stand to win over computerized customers. This month, OpenAI released Agent, an automated assistant that can search the web and take actions on behalf of consumers. Meanwhile, Microsoft launched “Copilot Mode” in its Edge web browser, an AI assistant for navigating the internet, and AI startup Perplexity is developing a browser called Comet that has e-commerce integrations. Just last week, Walmart gave a presentation of four “super agents” including one called Sparky, which was built to assist shoppers.
The field is still nascent but growing, and development of AI agents is expected to attract about $47 billion in investment in 2030, a tenfold increase from this year, said Nicole Greene, VP, analyst, market intelligence firm Gartner. With AI agents gaining momentum, brands need to be ready. Here’s what they should know.
How does agentic AI influence e-commerce?
The agentic movement will have significant repercussions as more people use AI agents to search for what to buy and book online, said Todd Parsons, chief product officer at e-commerce media platform Criteo. “Overnight, shopping has gone from people hunting for products to personal AI assistants hunting for answers,” Parsons said.
Brands are already overwhelmed by the retail media landscape, where every website becomes another advertising expense, said Jeriad Zoghby, chief commerce strategy officer at Interpublic Group of Cos. Brands have been dedicating more of their media budgets to retail media as retailers, including Walmart and Amazon, create more sophisticated digital ad platforms. Brands also have to consider social media and connected TV, and now generative AI is opening yet another channel, Zoghby said.
Last week, IPG launched what it called “Agentic Systems for Commerce,” a platform for brands to manage their digital shelves. The idea is to use AI to help track products, sales volumes, competitors’ products, media and other variables.
“Once you start seeing digital commerce as a thing you can control versus something you can’t keep up with,” Zoghby said, “it changes the whole business.”
How can brands communicate with AI agents?
Brands should apply some of the same lessons they have learned about answer engine optimization to communicating with agentic AI, Cadwallader said. For instance, it’s still important to create content for websites that agents can easily read by deploying common AEO practices, such as writing in frequently-asked-question formats. “Brands need to create content that is prepared for AI rather than human beings, as AI is now browsing on behalf of humans,” Cadwallader said.
Still, there are differences between how ChatGPT crawls a website to retrieve responses for its users and how it scans a website as an autonomous agent, Cadwallader said. ChatGPT’s agent scans the entire page, as if it were viewing the actual website, including the advertising, which is an interesting development regarding the types of content it ingests, Cadwallader said. The content of the ads served to a person’s browser could provide context to the AI about the interests of the consumer.
Brands also need to develop ways for agents to use the tools that consumers otherwise might use, especially software platforms, such as Hubspot and Salesforce, Cadwallader said. If an agent can interact with enterprise software tools on a platform, the agent will continue to visit those sites when it needs to perform its tasks, Cadwallader said, and the same factors could apply to more consumer-focused brands that offer online ordering and other services.
“It’s going to become very important that these agents … don’t just understand your business,” Cadwallader said, “but they can actually use your tools, as well.”
Should brands develop AI agents?
Brands need to build their own agents that can live on their websites and be at the ready to talk with the third-party agents, said Simon Poulton executive VP, innovation and growth, at e-commerce platform Tinuiti. Microsoft, for instance, is helping brands that Tinuiti works with enable agent-to-agent communication, Poulton said.
David Jones, CEO of the Brandtech Group, the AI ad agency company, is working with brands to prepare them for agentic AI. OpenAI is offering the building blocks for brands to build for the autonomous agents by developing its application programming interfaces (APIs) and an Agents SDK (software developer kit), Jones said.
“We’re already using agents as part of audience planning,” he said. “Brands will not only need to reach and influence people, but also the [large language models] and AI agents people use on their behalf.”
One of the considerations brands have to make is about how open they are to bot traffic, which is typically a dirty word in advertising. However, agentic traffic is automated, and brands need to rethink their bot-blocking policies, Jones said. “Many brands will have a lot of basic technical work to do to ensure they are not limiting their AI visibility,” Jones said.
Agents are still limited, however, and they are not handling transactions autonomously, Poulton said. Brands just need to prepare for the day the agents do get control of the credit card. “AI agents live on a spectrum,” Gartner’s Greene said. “Current agents behave more like assistants or copilots. As they improve, they’ll become more independent, acting more autonomously.”
