When AI does the shopping, how does your brand win?

As AI agents take on more of the shopping process, brands face a new question: how do you win a customer when an AI agent is doing the comparing, choosing and buying?

Agentic commerce is a form of online commerce in which AI agents research products, compare options and, where authorised, complete purchases on a customer’s behalf. While this kind of shopping isn’t yet mainstream, developments from Google, American Express and other tech companies show that the technology and infrastructure are beginning to take shape.

Imagine a family planning a holiday. They tell an AI assistant what they want: somewhere warm in April, suitable for two adults and two children, with a pool, direct flights and a budget of £3,000. The assistant researches destinations, weighs up hotels and flights, and draws up a shortlist. It might even go as far as booking the trip.

The challenge for brands is to ensure AI agents can find their products, recognise their value and complete purchases on customers’ behalf, wherever they shop.

How AI shopping agents are changing e-commerce

On 15 October 2026, Google launched Universal Cart in the US, as reported by TechCrunch. It lets shoppers collect products from Google Search and Gemini, track prices and receive restock alerts. It isn’t an autonomous shopping agent, but points to a future where shopping increasingly happens across platforms rather than on a single retailer’s website.

Google is also building the infrastructure to let AI agents complete purchases. Its Universal Commerce Protocol (UCP), introduced in January 2026 with retail and technology partners, is designed to help agents and businesses handle everything from product discovery to checkout and after-sales support. In May, Google announced plans to extend UCP-powered checkout to Canada and Australia, followed by the UK.

If an AI assistant is comparing offers, will customers stick with a familiar name or choose a competitor that better fits their needs? Reputation and customer experience will still count, but brands must also ensure agents can assess their offers and complete purchases.

Can AI shopping agents access your website?

There’s a catch, though: not every business wants AI agents accessing its services, let alone buying things on a customer’s behalf. On 17 October, The Verge reported that websites are blocking AI agents, with concerns over security, licensing and control, prompting companies to restrict access.

If an agent can’t access reliable product information, check prices or complete a purchase, it may be unable to recommend a brand, however well suited its products might be. Businesses need to enable legitimate agents to access up-to-date information and interact with checkout and customer services securely, while protecting customer data, commercial interests and control of the customer relationship.

Trust and security in agentic commerce

Even if an AI agent can find the right product and access the retailer, it still needs permission to spend the customer’s money. How can the customer be sure it will act on their instructions?

On 16 October, Digital Commerce 360 reported on American Express’s Intent Intelligence framework, designed to record and verify what a customer has authorised an AI agent to buy, including spending limits and merchant preferences. The aim is to help establish whether an agent acted within those instructions if something goes wrong. The framework had been published to encourage wider adoption but had no live integrations at the time of the report.

What if the price changes, a booking has the wrong cancellation terms, or an agent buys from an unauthorised seller? Customers need confidence that agents will follow their instructions and that problems can be put right.

Trust will vary between markets, shaped by local expectations, payment habits and attitudes towards AI making decisions. International brands need accurate product information, clear pricing, secure payments and helpful customer support, adapted to local expectations.

It also needs to be clear who is responsible when something goes wrong. That might be the retailer, the AI provider, or the payment company, depending on how the transaction works. Brands need to understand where responsibility sits and what happens if a purchase doesn’t go to plan.

Agentic commerce across global markets

At Oban, a key question is whether an AI agent that works in one market will work the same way in another. Discoverability is only the first step: agents must also be able to access local products, compare prices, use available payment methods, and navigate delivery and returns.

An assistant might understand the same request in the UK, US, Germany, Japan and Brazil, but the options available to it will vary. Currency, shipping costs, product ranges and customer expectations can all affect competitiveness and whether a purchase can be completed.

Google’s Universal Commerce Protocol (UCP) aims to make it easier for AI agents and businesses to work together, but a shared technical standard can’t guarantee that every retailer or payment service is available in every market. Its rollout depends on participating merchants and partners.

Global brands need to understand where these gaps exist, which markets they affect and what needs to change before AI agents can complete a purchase reliably.

Oban’s Local In-Market Experts (LIME) network can help you assess how well AI shopping agents work across different markets. Local specialists can give agents the same shopping brief in several countries and compare whether they can find suitable products, access accurate prices, use available payment methods, navigate delivery and returns, and complete the transaction. Testing real shopping journeys can reveal market-specific barriers that a global assessment might miss, helping brands prioritise investment.

How brands can prepare for agentic commerce

Businesses don’t need to overhaul their digital infrastructure immediately. Start by assessing what customers might delegate to AI agents and what could prevent your business from participating. Three areas are worth exploring:

1. Make your offer easy to understand and verify. Review product information, pricing, availability, delivery, returns and other purchase conditions. Check where this information is published, how often it changes and whether authorised third parties can access it.
   
   
2. Examine the route from recommendation to transaction. Map the journey from product discovery to purchase and after-sales support. Which steps can an agent complete, where does it encounter restrictions, and what systems, partners or permissions are needed? Focus on routes that are commercially worthwhile and secure.
   
   
3. Test the experience market by market. Don’t assume a capability available in one country will work everywhere. Test consistent shopping scenarios across priority markets, using local expertise to identify differences in availability, language, payments, delivery and customer expectations. Use the findings to prioritise investment.

Measuring success in agentic commerce

Agent-led purchases may create a measurement challenge beyond AI search visibility. If an agent completes a transaction without a conventional website visit, brands need to understand which systems are generating sales, how those transactions are recorded and whether existing analytics can capture them. As agentic commerce develops, businesses should identify the data they can access and the gaps they need to address.

So what does agentic commerce mean for your brand?

AI agents won’t replace the things that make people choose brands: good products, fair prices, reputation and service. But as agents take on more of the research and buying, brands need to understand what makes their offers competitive, whether agents can complete transactions and how these capabilities vary between markets. Get in touch to explore what agent-led shopping could mean for your business.