Amazon's Rufus and the limits of AI-assisted shopping

Happy Tuesday. I’m trusting that some of you are already taking some well deserved downtime to refresh before the New Year. I’m going to try to take a couple of days myself, but still plan on sending a newsletter next week because The Aisle still feels way too young for a whole week off. Fellow founders can let me know how long it’ll take for me not to feel that way, but I’m pretty confident it’s going to be a while! I’m OK with that.

Separately, thanks to those of you who’ve already replied to my weekend newsletter requesting your predictions on how AI will or won’t enhance or disrupt commerce in 2026. Please keep them coming. I’m learning a lot and will plan to include some in a future post so I can share all the knowledge. A reminder that they needn’t be long or super formal; a couple of sentences is just fine.

Lastly, something I’m keeping an eye on going into 2026: how much retailers, brands, and online marketplaces decide to lean into AI-powered dynamic pricing, personalized pricing, or customized pricing in 2026.

This was a topic already on my radar but was brought top of mind when Consumer Reports reported earlier this month that Instacart was allowing its retail partners to “charge up to 23% more for the same item ordered from the same store at the same time.” (Seems like some good ammo for the AI-enabled grocery pricing startups that I wrote about a few weeks ago that are pushing for more pricing transparency.)

Amid the backlash, Instacart announced on Monday that the company has ended such practices, even while quibbling over the phrases dynamic pricing and surveillance pricing saying that their tests were “never based on supply or demand, personal data, demographics, or individual shopping behavior.” OK.

Still the shift is a noteworthy one; just a few day ago, I was tipped off to some young consumer brands taking similar approaches. I also was chatting recently with leaders from an iconic fashion brand who were debating the ethics of such practices. I’m curious: are any of you experimenting in this space or have strong opinions about marketing different prices to different customers? Let me know. And yes, I’m happy to discuss topics like this one off the record.

Now on to the good stuff.

The first time I heard about Amazon’s Rufus shopping assistant, it wasn’t public and it wasn’t yet called Rufus. It was early 2023 when I learned that the tech giant was recruiting some former journalists to build teams that would train an AI shopping assistant to sound more human.

Fast forward about a year to early 2024, and Amazon announced the launch of Rufus, “a generative AI-powered expert shopping assistant.” (Former journalists are indeed among those working on the Rufus team at Amazon, which I admit is maybe only interesting to this journalist right here.)

Less than two years later, the company says ol’ Rufus is on track to generate $10 billion in incremental annual sales, according to some not-totally-clear internal methodology. (Have thoughts about this figure and how Amazon got there? I’ve gotten some intel already but am all ears for more.)