Morrisons ends AI smart trolley trial in Preston as Instacart retires carts

What happened

Morrisons has ended its trial of AI-powered smart trolleys in Preston, after its technology partner Instacart began winding down its cart operations. The supermarket had been testing the carts in one of its stores, and the technology is now being retired rather than rolled out.

TechRadar's coverage frames the outcome as a cautionary note for the wider grocery sector. Its headline argues that supermarkets are still pursuing automation, even though the trial's end suggests some of those bets may be aimed at the wrong technology.

The available reporting is limited on the specifics of the trial's design, its duration and the reasons behind the wind-down, so this briefing does not speculate beyond what has been confirmed.

Why it matters

In-store AI tools are often pitched as a way to speed up the shop and reduce friction at the till. A retired pilot is a useful reminder that a convincing demonstration does not guarantee a service model that works at scale. For leaders in retail, the question is whether the technology removes a real point of friction for shoppers and staff, or simply adds a new layer of hardware and software to manage.

The story also highlights the dependency risk in partnership-led digital transformation. When a retailer relies on a third-party operator for a customer-facing system, changes on the partner side can end a programme abruptly, leaving the retailer to explain the outcome to customers who may have noticed the change in store.

The Renascence take

Retailers are under pressure to show visible innovation, and smart trolleys make a compelling picture. Most readers will see a technology story, but the real lesson is about the shopping journey. Before automating a step, operators should map what the shopper is actually trying to do at that point, and whether the tool shortens that task or just changes who performs it.

The shopper's problem is rarely the trolley itself; it is the trip from aisle to checkout. Any AI pilot that does not measurably reduce effort at that moment is a demonstration, not a service improvement. Customer-obsessed operators should test the job first, the technology second, and set clear success criteria before a trial begins. They should also plan for the partner walking away, because in a pilot, that is when the real experience design gets tested.