How shopping chatbots might transform retail

Retail industry giants are taking different approaches to a technology that could further automate shopping

The advent of agentic AI threatens to start unpicking an online retail ecosystem that has cost billions of dollars of investment to create. Photo by Blue Planet Studio/Getty Images/Postmedia files

Article content

Some of the biggest crowds at Retail’s Big Show in New York this month were drawn not by retailers but tech executives. Groups such as Google, Microsoft and Salesforce erected stands the size of houses, while an advertising truck cruising the avenue outside the convention centre summed up the zeitgeist, its backlit cargo box promoting “AI chat that’s meant to deliver.”

Article content

Article content

Bots are learning how to shop. Sales driven by AI platforms will account for about 1.5 per cent of United States retail e-commerce this year, according to a forecast by research company Emarketer, but the potential impact of the technology was the dominant topic of conference conversation.

Article content

Article content

Article content

“For years, online shopping has been about keywords, filters, drop-down menus. And scrolling through multiple pages (of search results) until you find what you want,” Google chief executive Sundar Pichai, who was joined on stage by Walmart’s incoming boss John Furner, told his audience at the show. “Now…AI can do the hard work.”

Article content

Its promoters say that so-called agentic AI could be a step forward similar in significance to the start of online shopping in the 1990s, or the advent of smartphones in the 2000s, as autonomous agents cut out the tedium of searching and comparing.

Article content

Start-ups including Perplexity and OpenAI, as well as tech giants such as Google, want the tools to be the first port of call for purchasing goods, making AI-powered chatbot platforms the new storefronts of online shopping.

Article content

Article content

Fidji Simo, a former Instacart chief executive who is now OpenAI’s head of applications, has heralded the opportunity the technology poses for users’ quality of life. “We have a chance to make these time-saving capabilities feel not only useful but routine,” she wrote on OpenAI’s blog last year.

Article content

Article content

Yet the advent of agentic AI also threatens to start unpicking an online retail ecosystem that has cost billions of dollars of investment to create. Control over customer relationships and data troves that have been built up over decades could be partially ceded to intermediaries.

Article content

Replacing human eyeballs with digital ones risks undercutting lucrative “retail media” businesses, whereby retailers sell enhanced positioning in search results and advertising banners to vendors in a global industry that research group WARC estimates is worth nearly US$200 billion a year.

Article content

Some cybersecurity experts warn that handing over personal data to shopping bots will invite fraud, while others caution that they could be led astray by cleverly designed website code. There is also a risk that shoppers will choose not to outsource what for many is an enjoyable pastime, in the same way that they have not abandoned shopping in physical stores.

Article content

Amazon, the company that did more than any other to drive e-commerce adoption, has been notably more defensive than Walmart regarding AI, even though its web services division is pouring billions into the infrastructure needed to support large language models. Its website blocks many of the chatbots developed by start-ups and tech groups — including Anthropic, in which it owns a stake.

Article content

The group, which on Wednesday said it would cut thousands of jobs in order to reduce costs, is in talks with OpenAI about a US$10-billion investment. That may be accompanied by some form of commercial relationship involving its marketplace, but analysts expect any agreement will be more restricted than the partnership that OpenAI and Walmart unveiled last year.

Article content

Amazon’s chief executive, Andy Jassy, has also assured shareholders that it is “having conversations” with third-party agents and expects it will partner with some of them over time. “It reminds me in some ways of the beginning of search engines,” he told investors on an earnings call in October. “You had to kind of figure out the right way to work together.”

Article content

Article content

The company also said in a statement that it sees “significant opportunities to innovate and further improve the online shopping experience for customers as both an online store and agentic AI shopping assistant that shops other stores across the web.”

Article content

James Cadwallader, chief executive of Profound, a start-up that works with brands to optimize their AI reach, describes AI adoption as “a one-way door,” using a term for big strategic decisions popularized by Amazon founder Jeff Bezos.

Article content

“Agents have the potential to be existential,” he adds. “There’s a reluctance from platforms like Amazon to dive into things headfirst and hand over the keys to the castle.”

Article content

Article content

Since the 1990s, e-commerce has entailed typing and navigating around a search engine or a retail website to compare products, before placing them in a digital basket and entering delivery and payment details.

Article content

Using agentic AI, the shopper might enter a more stream-of-consciousness desire: “Buy me an insulated waterproof jacket for under $150, size women’s medium, for a Niagara Falls tourist boat ride in early March.”

Article content

Article content

AI would then scrape the internet for information on jackets, including product reviews and weather records from the Falls. Combining this with the shopper’s preferences from previous encounters, or following a dialogue, the chatbot would present a set of options. If authorized, it would then purchase the jacket without the consumer ever leaving the portal.

Article content

Amazon, Walmart and other large retailers have already deployed digital assistants with some AI capabilities. “Sparky” answers queries from Walmart shoppers about products in a conversational manner and learns from previous interactions, for instance.

Article content

Amazon’s equivalent, Rufus, offers functions such as “Buy for Me,” which enables users to find and purchase products from unlisted retailers through the company’s app. It is also experimenting with sponsored prompts that direct users towards particular products. The company said Rufus was “rapidly expanding both in capabilities and customer adoption.”

Article content

But Walmart, which has spent the past decade turning its e-commerce operation into a sales channel that is expected to bring in almost US$140 billion of revenue in the year to January, is going further. This month it hired Google and Amazon alumnus Sean Scott as senior vice-president of AI shopping.

Article content

It has also agreed to a partnership with Google that will surface the big-box chain’s products when users search using Google’s Gemini AI chatbot. Google has similar agreements with other retailers and marketplaces, including Etsy, Target, Shopify and Wayfair.

Article content

Furner told the New York audience that agentic AI “is really significant, just as search was for e-commerce a few years ago. It changed the face of commerce … I think the same thing is going to happen with agent-led commerce.”

Article content

In October, Walmart struck a deal with OpenAI, the company behind the ChatGPT chatbot, to power its drive into AI-powered shopping.

Article content

Hari Vasudev, chief technology officer at Walmart’s U.S. division, says agentic AI agents will reduce customers’ “cognitive load,” adding that it is working to make Sparky capable of understanding not only text inputs but voice, image and video.

Article content

Vasudev says a parent planning a child’s birthday party will “just tell Sparky: ‘Hey, I’m trying to plan a birthday party for my daughter, who’s 10 years old, and this is a theme that I would like, and here’s an approximate budget, and the party’s going to be on this date.'”

Article content

Article content

“And over time … the agent might actually complete the purchase on your behalf, and things will just show up at your home,” he adds.

Article content

According to data from analytics firm Similarweb, Walmart’s efforts have helped to double its share of referrals from AI, measured against a sample of major retailers and marketplaces, to 32.5 per cent over the course of 2025.

Article content

By contrast, Amazon.com’s share of referrals dwindled from 40 per cent in January 2025 to less than 11 per cent by December, though it remains the leading marketplace in terms of engagement both in the app and on the web and there is little evidence that customer numbers or sales have suffered as a result.

Article content

“Rufus captures demand but it doesn’t drive engagement,” cautions Hugh Hinkson, chief commercial officer at Pattern, an e-commerce accelerator and one of the largest third-party sellers on Amazon. “It cannot compete with agents that exist outside a single marketplace.”

Article content

Hinkson adds that Amazon’s chatbot block is detrimental for vendors listing goods on the platform who would welcome being reached through another portal. “It doesn’t benefit anyone to push away free traffic,” he says.

Article content

Article content

Perplexity’s Comet AI browser agent sidestepped the restrictions, leading Amazon to sue in November, alleging that the start-up is violating its terms of service and claiming that its agent degraded a customer shopping experience it had spent decades investing in.

Article content

Perplexity said in a statement at the time the lawsuit was a “bully tactic” and that its AI agents were a legitimate proxy for users.

Article content

Article content

For retailers, the potential of agentic e-commerce has to be set against both its investment requirements and its impact on marketing channels and revenue streams that have been built up over the best part of two decades.

Article content

Retail chains and online marketplaces have spent years perfecting ways of maximizing their presence in Google’s search algorithm, including paying for preferential placement.

Article content

More recently, they have turned to social media platforms such as Instagram, Facebook and TikTok to boost awareness and sales, often via influencers who have large numbers of highly engaged followers.

Article content

AI agents will also need to win over consumers skeptical that the jacket or birthday party materials the chatbot orders are indeed tailored to them, rather than driven by a retailer’s internal sales targets.

Article content

“You might expect that the chatbot system will give you an unbiased recommendation, but it might give you an advertisement without disclosure,” says Daniel Kang, assistant professor of computer science at the University of Illinois Urbana-Champaign.

Article content

Fewer than half of consumers reported trust in AI during the holiday shopping season, according to an Adobe survey, though the share is rising.

Article content

Kayne McGladrey, a senior member of IEEE, a technical professional organization, warns that AI agents raise serious concerns about mass surveillance, fraud and security.

Article content

One problem is that agentic AI reads all the text that it encounters and retains the data it absorbs, McGladrey adds. Embedded text, contained in website code but not visible to the human user, can trick agents into purchasing unwanted products while clones of legitimate retail websites can extract customer payment credentials.

Article content

Article content

“There are absolutely no scams on the internet, and everything is true on the internet,” McGladrey adds sarcastically. “And so this obviously will work really well with large language models.”

Article content

Retailers retort they have no incentive to mislead shoppers. “For 45 years we’ve been trying to build trust with the customer,” says Jordan Broggi, president of online at Home Depot, the largest U.S. home improvement chain. “That to us is much more powerful to protect than, like, ‘Let me sell you one more thing with an AI bot’.”

Article content

For retailers with robust e-commerce operations, agentic chatbots could curtail the budding “retail media” businesses that sell advertising exposure to vendors. If the shoppers are chatbots and not humans, will they be swayed by adverts, or trained to ignore them?

Article content

Google has introduced personalized ads as part of the company’s AI mode, a chatbot feature integrated into the company’s search results. OpenAI is also experimenting with ads on its chat feed. The financial terms of these have not been disclosed.

Article content

“If you’re taking consumer activity and moving it from a retailer’s website to a third-party environment, a search ad could still exist on the retailer website,” says Sarah Marzano, an Emarketer analyst. “But its value is inherently eroded because you’re going to have fewer consumers seeing those ads.”

Article content

Article content

This question is particularly acute for Amazon, which has harnessed its dominant e-commerce position to capture nearly 80 per cent of retail media digital ad spending, Emarketer estimates.

Article content

However, retailers have other features that could help them retain ownership of their customers. One is loyalty schemes such as Amazon’s popular Prime subscription, which offers faster delivery and other perks in return for an annual fee.

Article content

“It’s that spectre of, ‘Will I get dis-intermediated? How do I ensure they will still come to me?'” said Petra Schindler-Carter, general manager of retail and consumer goods at Amazon Web Services, at a New York media briefing. “That comes back to loyalty, and what is the reason you give them to come to your site.”

Article content

And while agentic AI could control more of the buying decisions, it cannot do the hard work of actually supplying and delivering the goods. The logistical power of Amazon, Walmart and others, including vast networks of warehouses and trucking fleets built up over several decades, means consumers can order merchandise and have it delivered or collect it quickly and reliably.

Article content

Neither OpenAI, Google or any of their rivals have signalled a desire to enter this arena with their own agentic e-commerce platforms.

Article content

“The role of Google in the middle of all of this is that of a matchmaker,” Vidhya Srinivasan, vice-president of Google ads and commerce, told another packed audience at the Big Show. “Not to be a marketplace.”

Article content

But for all the potential pitfalls, agentic AI is already sparking a gold rush. Search engine optimization, the mix of art and science that secures better placement on Google search result pages, is giving way to variations such as GEO and AEO — generative and answer engine optimization.

Article content

At Retail’s Big Show, a crowd formed around a salesman from Adobe as he demonstrated how to use its LLM Optimizer to enhance retailers’ visibility in AI-powered searches.

Article content

Google’s own business model is being disrupted by AI but its chief executive, Pichai, reminded his audience that past technology shifts “have brought new waves of innovation and opportunities,” describing AI as “the most profound platform shift of our lifetimes” and “an incredible accelerant of human ingenuity.”

Article content

Max Sinclair, who held several senior roles at Amazon before founding AI marketing platform Azoma, says the e-commerce industry has spent 10 years “optimizing for super-incremental improvements” in search.

Article content

“There’s now a space to reinvent online shopping.”

Article content

Additional reporting by Cristina Criddle in San Francisco

Article content

© 2026 The Financial Times Ltd

Article content

Article content