Latest
Deal Heidi cuts headcount by 18% and co‑founders leave after $475 million raiseDeal What buyers look for when they acquire an AI companyGlobal Which business processes can AI agents handle todayGlobal AI agents in accounting and finance teamsGlobal AI agents in logistics and supply chainsUnited States Texas elects first Democratic governor in decades, putting AI data centre boom at riskGlobal OpenAI launches GPT‑6 Cyber model for defensive cybersecurity
United States edition
Friday, 6 November 2026
Infercline.
Advertise
The AI Opportunity Desk

AI agents in retail and e-commerce

Retailers use agents to answer shoppers and run the back office, and shoppers are starting to send agents to buy for them. What is real, what is early and what the rules say.

The Infercline Desk · Friday, 6 November 2026 Sign in to save
a window with a sign and a guitar in it
Photo by Yehor Litsov on Unsplash

Retail and e-commerce meet AI agents from two directions. On one side, retailers use agents to serve customers and run their operations: answering questions, handling returns, updating product listings and keeping stock records straight. On the other side, shoppers are beginning to use agents of their own, asking an AI assistant to find and even buy a product for them. Both matter, and the second could change how customers find your shop at all.

Customer service is the most established use. Order tracking, delivery questions, returns and product questions are frequent and mostly routine, which suits an agent connected to your order system. The results can be large, but so can the risks. Klarna, the payments and shopping company, said in February 2024 that its AI assistant handled 2.3 million conversations in its first month, two-thirds of its customer service chats. By May 2025, Computing reported that Klarna was hiring human service staff again after its chief executive said quality had fallen. Speed and volume are not the whole story; customers still need a person when things go wrong.

Customers' attitudes deserve respect. A Gartner survey published in July 2024 found that 64 per cent of customers would prefer companies did not use AI for customer service. For a retailer, that argues for an agent that is honest about being AI, settles simple questions quickly and hands over to a person without friction.

Behind the scenes, agents can write and update product descriptions, sort supplier emails, flag stock that is running low, check prices against your rules and prepare reports. These jobs are less visible but often save more time, and a mistake is easier to catch before it reaches a customer.

The bigger shift is agents that shop. In April 2025 Visa announced Visa Intelligent Commerce, offering builders of AI agents a way to make payments with tokenised cards and spending limits set by the consumer. Mastercard announced Agent Pay the same month, built on what it calls Agentic Tokens. In September 2025 Google announced the Agent Payments Protocol, an open standard for agent-led payments that uses signed mandates as proof of what the user asked the agent to do. In September 2025 Stripe said ChatGPT had launched Instant Checkout with the Agentic Commerce Protocol, developed by Stripe and OpenAI.

This market is moving and not every launch has stuck. In March 2026 Digiday reported that OpenAI was moving Instant Checkout into apps, while Shopify said that millions of its merchants could sell to ChatGPT users through what it calls Agentic Storefronts. The direction is clear even if the details change: some shoppers will find and buy products through AI assistants rather than search engines or your website. Retailers need product data that an agent can read and trust: accurate prices, stock levels, descriptions, sizes and delivery times.

Reviews are an area where AI and the law collide. In August 2024 the US Federal Trade Commission announced a final rule banning fake reviews and testimonials, explicitly including AI-generated fake reviews. In the UK, the government said fake reviews were banned from 6 April 2025 under the Digital Markets, Competition and Consumers Act, with the Competition and Markets Authority given direct enforcement powers. Using an agent to write reviews, or to post them under false names, is a legal risk in these markets and a reputational one everywhere.

Customer data rules apply across all seven Infercline editions. Shopper names, addresses, order histories and payment details are personal data under UK GDPR, South Africa's POPIA, Nigeria's Data Protection Act, Australia's Privacy Act, Singapore's PDPA, the UAE and Saudi data protection laws and US state laws. Any agent that reads them must be set up to meet those rules, including where the data is stored.

Why it matters to your business

Retail margins are thin and customer patience is short. Agents can cut the cost of service and back-office work and help small teams keep up with larger rivals. At the same time, agent-led shopping may change how customers discover products, rewarding retailers whose product information is complete and accurate. Retailers who prepare now will be easier for both customers and their agents to buy from.

What to do on Monday

  1. Sort last month's customer contacts by type and identify the three most common routine questions for an agent trial.
  2. Audit your product data for accuracy: prices, stock, sizes, delivery times and returns policy. Fix gaps before any agent, yours or a shopper's, relies on it.
  3. Ask your e-commerce platform and payment provider what support they offer for agent-led checkout, and when.
  4. Confirm in writing that no one in your business or agency uses AI to write or post fake reviews.
  5. Make sure every customer conversation with an agent says it is AI and offers a route to a person.

More on this topic: AI agents by industry

Sources