
Critical Theory in AI
Screen economies and AI-powered price tags
AI-powered Electronic Shelf Labels are being adopted by major retail chains. These tiny screens turn the supermarket aisle into a real-time market.

The supermarket has always been a sensor. The point-of-sale system has logged every basket for thirty years. The loyalty card has matched those baskets to households for twenty. The shelf, however, has been mute. It has displayed yesterday's price in yesterday's ink. With AI-powered electronic shelf labels, the shelf becomes a transmitter — and the supermarket becomes, for the first time, a two-way conversation between the store and the shopper.
Prices that used to settle into the week now flicker by the hour. The bag of rice that cost one figure at nine in the morning costs another at five in the afternoon, because the model has noticed a weather pattern, a competitor's promotion, an inventory drift in the regional warehouse. The shopper enters a small, continuous auction without having agreed to bid.
A screen economy is not a new economy. It is the old one, accelerated past the speed at which we used to feel it. Dynamic pricing has existed since the first market. What is new is the latency. The market used to clear once a day. Then once an hour. Now once a minute. Below a certain threshold, the human cannot perceive the clearing, and the asymmetry between the algorithm and the shopper becomes structural.
There is a fairness question that the industry has been careful not to ask out loud. If the price on the shelf is a function of the model's belief about the shopper standing in front of it — and the shelf has cameras, and the cameras feed a re-identification system, and the system knows the shopper's loyalty history — then the price is no longer a property of the product. It is a property of the relationship. Two people standing in the same aisle can be quoted different prices for the same tin.
Regulators have been slow. The legal framework for price discrimination was written for an era in which the price tag was a physical object updated weekly by a teenager with a sticker gun. It does not contemplate a label that changes its mind between the time you reach for the shelf and the time you scan at the till. Several jurisdictions have begun to draft updates. None has finished.
There is also a labour story here. The teenager with the sticker gun was a job. The maintenance of the electronic label network is a different job, requiring different skills, located in different places. The transition has been presented as automation. It is more accurate to call it relocation — the work has not disappeared, it has moved closer to the warehouse and further from the floor.
For Caribbean retailers the question is whether to adopt the technology at all. The capital cost is non-trivial. The trust cost — in a region where the shopper-shopkeeper relationship is still a meaningful social bond — may be higher. A chain that flips its shelves to dynamic pricing overnight may discover that its customers preferred the stickers, and the teenager, and the small ritual of knowing what things cost.
— Fazal Ali · 18 April 2026 —
Next entry · Critical Theory in AI
The 100-Model War and shadow AI
The Hundred Model War started in 2023. From Beijing to Shenzhen, China is implementing its frontier strategy at industrial scale.
