New Bain report suggests three ways retailers can approach agentic commerce

“Retailers will need to rethink operations, marketing, technology, data, and pricing to stay relevant in an agent-first world,” write Bain partners Aaron Cheris and Justin Miller in the report developed alongside payments provider Stripe and shared exclusively with Retail Week.

Big moves by AI providers and retailers have forced the question of agentic shopping, which, in short, means AI bots that can search and even complete transactions on behalf of customers or brands. AI-driven transactions remain a small, but fast-growing slice of ecommerce. 

In January, Google announced its Gemini Enterprise for Customer Experience, a deployable shopping agent that brands can use to blend customer support and finding items. 

Last year, OpenAI announced its Agentic Commerce Protocol, a set of guidelines to help brands and developers set up retail experiences for AI agents. Several UK retail brands, including Frasers Group and JD Sports, have announced partnerships with Commercetools to set themselves up for agentic commerce. 

Separate research out from Adobe this week showed that 39% of UK consumers are open to interacting with a brand’s AI agent, while 27% don’t want to interact at all, with the remainder undecided. 

Build or embrace…

There are effectively three ways that brands can deal with agentic commerce, the Bain report says. The first is to “embrace the agents”, allowing third-party agents to crawl “their sites, list their products, and often close transactions”. This may suit brands that cannot drive traffic at scale or lack the resources to develop their own AI infrastructure 

More resource-heavy brands can follow the examples of titans like Amazon and Walmart in building their own agentic ecosystem. This “build the agents” approach could enable marketplace retailers to “make their brands synonymous with convenience and value.”

The interplay between these two approaches was highlighted by the rollout of Amazon’s ‘Buy For Me’ tool last year, an AI tool that allows users to buy products from other retailers if they are not available on the Amazon site. At the same time, Amazon has blocked bots from other AI sites, sending a cease and desist letter to AI-driven search engine Perplexity at the tail end of last year. 

The online retailer’s chief executive Andy Jassy said in an earnings call earlier this month that although it was likely that Amazon would have relationships with third-party platforms, retailers’ own agents will, in many cases, offer a stronger proposition. 

“These horizontal agents don’t have any of your shopping history,” he said. “They get a lot of the product details wrong, they get a lot of the pricing wrong. And so we have to try to find a customer experience together that’s better and a value exchange that makes sense for both parties.”

Content delivery network Cloudflare announced mid-way through last year that it would block AI crawlers by default, enabling websites to charge AI crawlers for access.

The charging landscape around third-party agents remains unclear. Although not quite agentic shopping, OpenAI confirmed in January that Shopify merchants would be charged a 4% fee if users transact through the ChatGPT platform. 

…or both

Cheris and Miller also suggest that a middle way is possible, with brands providing some exposure to third-party agents while also making some experiences only possible for shoppers using their sites directly. 

They give the example of the US brand Home Depot’s AI companion Magic Apron, which is only available on the brand’s website and provides specialised customer support not fully replicable on a third-party platform. 

This may provide some answer to the question of how to make a retail brand stand out if consumers are only interacting with it through an AI platform or, more to the point, agents are the ones interacting with it on their behalf. 

The Bain partners encourage retailers to own the checkout process where possible, seek partnerships to retain access to customer data, while also, unless they decide to stay fully closed to outside agents, optimising their sites in ways that “appeal to agentic logic.”