As these autonomous economic operators take on the shopping on behalf of the user, brands are finding they need to ask, ‘What do you do when your customer is an agent?’.
An AI agent is a system or program that can autonomously perform tasks on behalf of a user. In the case of shopper agents this means researching, comparing, and purchasing on behalf of customers.
Last year, Amazon launched its ‘Buy for Me’ functionality in Beta. This tool is able to order products on behalf of the user from other sites outside of Amazon’s own ecosystem. Since then, Amazon added an auto-buy feature to its AI assistant Rufus, allowing users to set a ‘right price’ for any particular item. When the product reaches that price, Rufus hits the buy button on their behalf.
Though these are limited use cases, they are real life examples of agentic commerce in the wild among consumers – outside of AI evangelists.
Consumers using agents to take the work out of finding the best deal, or the right product for them – at least in utility categories
Simple as these examples are, they do give a potential taste of what the agentic future may look like in the short-term. Consumers using agents to take the work out of finding the best deal, or the right product for them – at least in utility categories.
Though current agentic solutions to shopping do require users to make a specific request, platforms are testing systems that work based on intent. This means that Agents will be able to anticipate what the consumer wants, based on a rich profile of their preferences, behaviours and lifestyle. However, for consumers to hand over the full process (and their credit cards), with no oversight or final sign-off, will require significant trust in the technologies. Something the current breed of agents have not, as yet, potentially earned.
Infrastructure will of course be key in developing that trust, with tech platforms like Amazon, OpenAI and Google and financial companies like Mastercard and Visa already building the foundations required to allow these agentic dreams to become reality.
This matters for brands and retailers as that infrastructure will shape what they need to do to ensure they don’t miss out on agentic transactions. A key update brands and retailers should be aware of is the launch of Google’s Universal Commerce Protocol (UCP).
The UCP is an open-source standard for online shopping. It is essentially a shared set of rules and language that allows different ecommerce players to talk to each other without friction. Google is not alone in proposing a protocol designed to provide the structure AI agents require, but it does seem to be the one getting the most traction.
Adopting a UCP ensures information about products, baskets, checkout and orders is organised in a way that is same across different stores, platforms and apps.
Though this protocol covers all of ecommerce, it is designed to facilitate (and win Google a slice of) the agentic purchase journey. Having a consistent approach allows agents to understand and find products in a predictable way across different sites, removing the risk that the agent either ‘guesses’ wrong or ignores a site that it cannot understand.
Adopting a shared language is powerful. There are risks to organisations that whichever protocol they adopt leaves them too dependent on a particular platform. Something that whichever platform wins the race to become the standard, is likely aiming for.
While those platforms vie for position, the place where we will likely see widespread adoption of AI agents first is China. With the combination of a digital ecosystem dominated by super-apps, which already connect multiple services, along with robust authorisation infrastructure and the penetration of digital payment methods, China has the ideal foundations for agentic commerce to function.
Brands and retailers need to prepare for a future where their relationship with the user is mediated by an autonomous technology – at least for parts of the journey.
Agents shift the sphere of influence
Agent-mediated shopping will shift how and when brands can enact influence. For the human element of that influence, it likely moves up funnel. If the research phase, (seeking out the best deal and the actual transaction) is being handled by the customer’s agent, then the battle is really about being the user’s pre-existing or named preference.
This can be thought of as the difference between asking an agent to book ‘an Italian restaurant for Friday night’ and booking ‘My favourite Italian restaurant’.
This will have an impact on where marketers will want to place their focus and spend their budget, as performance ads lose their role and brand becomes only more important.
In a world of shopper agents, of course the human represents only one half of the sphere of influence. Agents can be influenced as well.
In a recent piece of research in the Harvard Business Review authors Mark J. Greeven, Fabrice Beaulieu and Wei Wei describe this as competing for the ‘agent shelf’, saying that ‘The decisive question becomes: Under what conditions will an agent reliably include you, even before a customer sees alternatives?’
For retailers and brands this means structured product data (ensuring the agent can see, understand and do business with you) and then reputation and trust signals. Which agents will likely use to ensure they are making the right choice for their user.
Just like with ecommerce, the role of shopping agents will likely be different depending on category. The way we purchase for a meat thermometer online is very different to how we shop for a car, for instance and the same will be true in a world where we all have a personal shopping agent at our disposal.
This means exactly how and how much of the customer journey agents will absorb will depend on your product, customer type and how customers view you. The best advice is then to stay close to customer behaviour and start building the data and reputation layer that will help you win, whether the final decision is made by a human or an agent.
For more on the rise of AI agents register for Econsultancy’s Digital Shift to hear the latest about AI’s role in CX. 23rd June 2026, 9am and 4pm BST
Or explore all of Econsultancy’s AI reporting and guidance here.