16/09/2026
For years, I've seen merchandising teams keeping the product catalogue working through what is effectively an invisible expert system.
It isn’t a platform or a dashboard. It’s the accumulated judgement of skilled people who know when a product title needs more context, when an attribute can be technically complete but still commercially misleading, or why two near-identical SKUs are not actually interchangeable.
They’re also the people checking that availability, variant logic, imagery, dimensions, pricing and product relationships all make sense before a customer sees them.
A lot of that decision-making happens in spreadsheets, PIM workarounds, trading conversations and last-minute feed fixes. Over time, that knowledge becomes something like the product catalogue’s hidden operating system.
As commerce becomes more agent-mediated, though, I'm seeing that retailers may have less opportunity to rely on people fixing incomplete or ambiguous product information by hand.
An AI shopping experience can’t draw on the distinction an experienced merchandiser carries in their head. It can only work with the data, relationships and rules it has been given.
Google’s latest work in ecommerce and agent protocols is one indication of where things are heading. For me, the interesting implication isn’t that merchandising judgement becomes less valuable. Quite the opposite.
The opportunity is to capture more of that judgement and turn it into durable, usable product data.
So perhaps the question isn’t simply whether your product catalogue is complete. It’s whether the knowledge that makes it genuinely sellable still lives mostly in the heads of the people who step in and repair it.
So, my question is ... If an AI shopping experience had to interpret your catalogue tomorrow, which product-data decisions would it get wrong because your best merchandisers currently fix them manually?