Getting ready for agentic commerce: Can an AI assistant tell you if your product is the right one?

It is Wednesday evening. You want to paint a 14 metre square bedroom in pure white (RAL 9010) before the weekend, starting from nothing: no paint, no roller, no tape, no sheeting. So you ask an AI assistant: please order everything I need so I can collect it asap.

The assistant doesn’t open six tabs, read through all websites and compare photos. It reads product data from every paint retailer and checks the coverage per litre, how many coats the walls need, which roller suits that paint and whether everything can be delivered in time.

That paint might be right and the roller too. But if the product data does not say the coverage per litre, or which roller is best for that paint, the assistant can’t do a decent job, so moves on to a retailer whose data does. 

In our first blog in our Retail Radar series on product life cycle management (PLM), we argued that trusted product information is a competitive advantage. This time we’re looking at agentic commerce and what it asks of the product catalogue.

Regulation is ahead of consumer awareness

Agentic commerce is a form of digital shopping, where consumers instruct AI assistants to independently research and compare products before making purchases, all without requiring manual browsing or checkout steps.

In January, Google and Shopify launched the Universal Commerce Protocol at NRF, which is backed by more than twenty commerce and payment partners. UCP is an open standard that lets AI assistants and webshops exchange information in a fixed way, so a customer can find, compare and buy a product through an assistant instead of on a website.

At the moment, it is only available in the US, but is expected to launch in Europe in 2027, once the rules under the Digital Markets Act have been sorted out. By 2030, it is estimated that 10 – 25% of all online shopping will be done by AI assistants.

But rather than waiting for Europe to catch up from a regulatory perspective, retailers need to start thinking about this now. Getting your data ready takes time, so an early start could mean establishing a competitive advantage or at least stop you from falling behind.

That readiness is not just for retailers’ own agentic capability. Google has also introduced a ‘business agent’ that a retailer configures with its own product data and content, like brand guidelines and size charts, to answer customer questions directly in the retailer’s tone of voice. Google’s business agent is already live in the US, while OpenAI is still testing this concept. And Meta has just launched ‘Muse’, which is apparently the world’s first personal AI agent.

The AI assistant doesn’t browse – it asks

At the moment, we type a few words into a search bar, look at the results, open three product pages and decide. But with agentic commerce, the customer starts with a requirement and hands it over to the assistant  (e.g. ‘I need a heavy-duty waterproof coat that will be good in sub-zero temperatures for my commute to work’). The AI assistant then reads the product feed, the catalogue in a form a machine can read, rather than the website. It compares products from different retailers on whatever data it finds there. 

In our first ‘paint a room’ example, the agent not only has to check the product data to ensure these items work together, it needs to check data sets like live-stock availability per store to make sure that same-day collection is possible.

Why customer complaints and returns are valuable

Product data quality needs to be measured differently: not just the list of attributes retailers choose themselves, but the questions customers actually ask, and whether the product catalogue gives an answer in a form an AI assistant can understand.

Retailers already have this information in the form of support tickets, reviews, search logs, FAQs, the reasons why customers send things back, etc.

Take colour. ‘White’ fills the field, but is it specific enough? RAL 9010, with the coverage per litre, gives a machine something to work with. It’s about having a data value that means something. All of these questions can help to inform the data the retailer needs to supply for an AI assistant to consider their product.

Gaps the Product Information Management (PIM) cannot close

When a question cannot be answered, the fix is not always a new data attribute: if a retailer wants to say how many litres of paint are needed to cover 14 square metres in two coats, and the supplier never recorded the coverage rate, no amount of work in the PIM will produce that answer. That is a PLM problem, not a PIM one: it has to travel back to the supplier.

There is also a regulatory consideration. The Digital Product Passport will soon require this same kind of checkable data. Batteries are first, from 18 February 2027, and most other products will follow within a few years.

Whether it is a customer question or new legislation, both can be checked with a scan against the product catalogue. Product data does not stand still, so this has to be a continuous loop – check the data, check compliance and feed back the information that falls short.

The action for retailers

In our first article we asked whether product data could stand up to scrutiny from regulators, AI assistants and customers. Agentic commerce makes that more specific: can an AI assistant tell, from your product data, that your product is the right answer to a customer’s need?

To help them answer ‘yes’, retailers have to do two things:

• Connect customer service data to product data: support tickets, chats, return reasons and reviews already show the questions product data fails to answer. Use that for data enrichment.
• Use the Digital Product Passport as the business case: the DPP asks suppliers for the same data on materials, origin and durability that an agent needs. Bring these two requests together and handle both in the same conversation with suppliers.

Agentic commerce means retailers need to think more laterally and proactively about their product data. And in a season where more customers than last year will ask an AI assistant to do their shopping for them, retailers need to get up to speed quickly.

Read more

Click here to read The Retail Radar, issue no. 1: “Is your product lifecycle ready for the future of retail?”

Get in touch

Do you want to understand what this means for your organisation? Bastiaan de Groot ([email protected]) and Kjeld Vermolen ([email protected]) help retailers create AI-ready product data, laying the foundation for agentic commerce, compliance, and growth. Get in touch to discuss how ready your data is for agentic retail.