AI shopping agents move from recommendations to the checkout

Agentic ecommerce is the latest stage in the long evolution of online shopping: software agents, powered by artificial intelligence, doing much of the work that people once carried out manually at a keyboard. Instead of simply recommending products, these tools can compare prices across sellers, watch for discounts, remember preferences, build baskets and even complete purchases, usually stopping short until a shopper gives approval.

The scale of the opportunity is one reason t...

he retail industry is paying close attention. McKinsey has estimated that by 2030, agentic commerce could generate as much as $1 trillion in orchestrated revenue in the United States’ business-to-consumer retail market, with global projections rising to between $3 trillion and $5 trillion. That kind of money helps explain why major technology companies, marketplaces and payment firms are moving quickly to build their own versions of the same idea.

At its simplest, the concept is a shopping assistant that sits inside a browser or app and does far more than earlier recommendation engines ever could. It can sift through product listings, weigh trade-offs, track changing prices and act on a shopper’s stated goals. In practice, that means the point at which a consumer forms an intent may become more important than the moment of checkout itself, because the agent is increasingly doing the comparison work beforehand.

McKinsey says this shift is not just about convenience; it also changes how brands and retailers influence demand. In luxury, for example, the consultancy argues that AI is beginning to mediate the consumer journey by interpreting, comparing and curating products on a shopper’s behalf, pushing decision-making further upstream and forcing brands to rethink how they shape desire in a more automated marketplace.

The technology did not appear overnight. It builds on earlier forms of conversational commerce, digital assistants, visual search and “buy for me” experiments. What changed in 2025 and 2026, however, was the infrastructure underneath. Google’s introduction of the Universal Commerce Protocol on January 11, 2026, signalled that agentic commerce was becoming a wider platform issue rather than a curiosity confined to a single app or pilot scheme.

Consumer attitudes suggest there is demand for help, but not necessarily for surrendering control. Gartner’s Kate Muhl said in a press release that shoppers are “not looking to outsource shopping decisions to AI”, but rather want assistance finding better information, comparing prices, identifying deals and narrowing options while keeping the final say themselves. That distinction matters, because it suggests the near-term winner may be the system that helps people decide rather than one that decides for them.

A Gartner survey of 322 US consumers in January 2026 found that 31% were comfortable letting AI narrow choices for household supplies, while 28% were willing to do the same for personal electronics. The numbers indicate cautious openness rather than outright trust. A separate Zeta survey of 2,000 US adults who use AI monthly found 70% liked the idea of receiving an AI-chosen gift, although 68% still felt gifts were more meaningful when selected entirely by a person.

Retailers and platforms have been racing to respond. Google’s holiday shopping update in November 2025 began rolling out agentic checkout and highlighted the scale of its Shopping Graph, which it said contained more than 50 billion product listings and was refreshed continuously throughout the day. On January 22, 2026, the company added more personalised recommendations in AI Mode, including shopping features designed to make product discovery more tailored to individual users.

Pinterest has also leaned into the trend. In late October 2025, the company upgraded boards into a more personalised shopping assistant and then launched Pinterest Assistant as an AI-powered, visual-first collaborator tied to its huge user base. The move reflected a wider industry view that shopping is becoming less about static search and more about guided discovery, with AI taking on the role of an interpreter rather than a passive tool.

Payment and checkout firms are moving too. Meta entered the space on September 8, 2026, with Muse, a personal AI agent designed to work across apps and the web inside a secure virtual machine. That same day, Stripe announced a Muse integration with Link, allowing faster checkout at more than 1 million businesses and making single-use virtual cards part of the process. A few weeks later, on September 29, 2026, Stripe said agentic purchases through Link had risen 38-fold over the previous month, a striking sign that the behaviour is moving beyond experiment.

For retailers, the implications are wide-ranging. McKinsey says agentic AI could free up as much as 40% of merchants’ time by taking over routine, repetitive tasks, allowing teams to focus on strategy rather than administration. That is a substantial productivity gain, but it also hints at a more competitive market, where product pages, pricing structures and fulfilment systems may need to be redesigned for machine readers as well as human shoppers.

For consumers, the short-term appeal is obvious: less time spent searching, more efficient comparisons and a better chance of spotting a bargain. The longer-term question is how much autonomy people are prepared to give up. For now, the evidence points to a cautious compromise: shoppers want AI to narrow the field, surface the best options and handle the tedious parts, but most still want to approve the final purchase themselves. Whether that line holds as the tools become faster, more persuasive and more deeply embedded in checkout systems may define the next phase of online retail.

Source: Noah Wire Services