According to Bloomberg Business, Instacart has ended an experiment where it used an AI tool to algorithmically set different prices for different customers. The program, which was detailed in a report this month from the Groundwork Collaborative, Consumer Reports, and More Perfect Union, resulted in shoppers being charged varying amounts for the exact same items from the exact same stores. The revelation came at a sensitive time, as inflation continues to drive up grocery costs. Following the report’s release and subsequent media coverage, including on Bloomberg’s Big Take podcast featuring antitrust reporter Leah Nylen and Groundwork Collaborative’s Lindsay Owens, Instacart made the decision to terminate the pricing experiment.
How the AI pricing worked
Here’s the thing: this wasn’t about dynamic pricing like you see with Uber Surge. This was algorithmic price experimentation. Basically, the AI was running constant A/B tests on live customers to see how much they could charge before someone noticed or balked. It’s a common tech playbook—optimize for maximum revenue per user—but applying it to essential goods like groceries is a whole different ballgame. The system wasn’t adjusting prices based on delivery distance or time; it was charging Customer A one price and Customer B another for the same box of cereal from the same digital shelf. And that feels fundamentally different, doesn’t it?
The broader context and risks
Now, Instacart is far from alone in using this kind of tech. Airlines and hotels have done it for years. But there’s a massive perception risk when you apply it to a basic necessity during a cost-of-living crisis. The backlash was swift because it taps directly into a fear of unfairness and a lack of control. Technically, the challenge is immense: calibrating an algorithm to extract value without triggering a PR disaster. The trade-off is between hyper-personalized pricing (which can sometimes mean lower prices for some) and the erosion of consumer trust in a platform’s basic fairness. Instacart clearly decided the latter risk was too great. So they pulled the plug. But you have to wonder, how many other apps are running similar tests right now, just under the radar?
