GUEST COMMENT How to succeed in the era of agentic commerce

James Taylor, CEO and Founder, Particular Audience, offers advice to retailers looking to boost their retail media monetisation capabilities

There’s a reason why Amazon is the world’s largest retailer. Anyone who has ever shopped there – and let’s face it, most of us have – will know that in addition to unrivalled choice for virtually any item on the planet, its search and recommendation engines, powered by AI, are best in class. 

One reason for this is because, when it comes to search, unlike the majority of retailers, Amazon doesn’t rely on exact keyword matching. Key in “anti-ageing cream” in to most cosmetics ecommerce sites and you would hope to get some returns, though perhaps not if you’ve used the US spelling “aging”. Put in a more conversational term, however, such as “how to reduce wrinkles” and Amazon will succeed in returning relevant products where many others fail. 

Both queries are looking for the same solution, but it’s beyond the capabilities of exact keyword matching to recognise that from any search term that isn’t an exact match for what’s in the retailer’s product catalogue. Exact-match keywords, in fact, miss 75% of intent. 

Amazon’s recommendation engine is similarly sophisticated. It doesn’t recommend by the segment it thinks you belong to, but by the intent you reveal as you browse the website – the search terms you enter; the product pages you look at before you put something in your basket. These real-time signals infer what you need right now, and are complemented by collaborative filtering, which uses the behaviour of people who bought what you just bought to predict what you might want to buy next. 

Many retailers fall into the trap of thinking that this degree of personalisation is not possible if you don’t have Amazon’s deep pockets. But this is a myth. Modular retail media platforms enable retailers to plug the gaps in what they already have, rather than rip and replace the whole stack. Personalisation is really just good prediction and Amazon-style personalisation is open to everyone, so long as you do the groundwork. 

So what do retailers need to do to up their personalisation game? 

The first task is to get your Model Context Protocol (MCP) architecture together; it’s a three-day project at most. MCP is an open-source standard that enables AI models to seamlessly connect with external tools and data. It allows retailers to build LLM integrations – known as apps – that customers can open through ads. In this way, ads in AI search become functional shopping experiences. A successful MCP deployment is key to retail media success in the AI era, but it doesn’t happen in isolation. It relies on clean attributes, reliable feeds and well-modelled commercial rules in the retailer’s catalogue. 

The MCP handshake – the process that establishes secure, standardised connections between AI models and external data sources or tools – is now being adopted widely across the AI industry as a standard for AI-tool communication.

It’s vital that retailers, not platforms, own this layer, in order to retain commercial control of their monetisation capabilities. Those that cede it – by exposing product catalogues directly to third-party LLMs without an intervening decisioning layer – risk becoming inventory in someone else’s auction.

How to own the decisioning layer

So, how do you own the decisioning layer? You need to move beyond exact-match keyword search, which is not fit for purpose in the AI era, and arguably never was. Vector/transformer semantic search delivers materially better relevance than keyword search, but needs a mechanism for merchandising control, margin prioritisation and sponsored activity. It’s strong for organic discovery and monetisation.

What retailers need to succeed in the era of agentic commerce is transformer search with a governance layer on top of it. This combines semantic ranking with merchandising rules, personalisation and sponsored margin opportunity to create a relevance curve fit for AI scrutiny. Whether retailers buy it or build it themselves, this is the architecture required to monetise AI-driven discovery. 

For retailers, the benefits come in the shape of higher-margin yield and monetisation potential. Advertisers, meanwhile, benefit from improved efficiency and validated ROAS through predictive targeting and automated campaign management. And let’s not forget customers, who benefit from greater relevance and utility. Product recommendations that match the intent they – or their agent – display when browsing a website. Ads that are relevant to what they searched for, rather than who bid most to take the slot.

In one test run with PetAds, an Australian retail media network operated by pet supplies retailer, PetBarn, switching from manual keyword bidding to automated coverage via transformer-driven search bidding boosted the proportion of performant search queries with monetisation coverage four-fold. 

That’s the sort of uptick no retailer can afford to ignore. 

Author

James Taylor is CEO and Founder, Particular Audience