During DoorDash’s Q3 earnings call earlier this week, CEO Tony Xu revealed that out of nearly 1 billion total orders in the quarter, third-party AI agent-placed orders accounted for just a few hundred.
Uber Eats did slightly better, with CFO Balaji Krishnamurthy reporting around 500 agentic orders out of its 320 million total food and ride orders during Halloween week. Just four of those were for actual rides.
Who were these people, we wonder? Was someone testing an integration, showing off to their date or simply unwilling to devote another 50 seconds of their life to opening an app?
The latest AI trends, brought to you by experts
Get curated insights on the most important—and intriguing—AI news. Subscribe to our twice-weekly Think Newsletter.
This tracks with something we’ve learned about AI: Rather than replace human tasks altogether, we let it handle the annoying stuff. DoorDash recently piloted an MCP connector with a few large tech companies that allows them to build shared tools around the ways their teams work—a Slackbot to collect orders, perhaps.
We love that a technology poised to revolutionize commerce might first solve the ancient workplace problem of getting 12 people to agree on lunch without having to watch Rudy in Biz Dev dismantle a tray of deli sandwiches and hurl all the bread in the garbage.
Speaking of things that perhaps don’t need highly advanced technical intervention, six robots have just completed a 24-hour shift in a convenience store, and two failed their performance reviews. Among their transgressions, a cleaning robot encountered a full trash can, wandered off its assigned route and left the store altogether.
The researchers behind the Agibot experiment explained that the hardest part wasn’t getting a robot to perform a specific task; it was getting the robot to respond and react appropriately when the real world threw it a curveball. We get it, seriously.
Meanwhile, food tech company Wonder is going all out on automation. Its new Infinite Makeline system can hold 68 ingredients and satisfy our insatiable desire for slop bowls, cranking out 500 per hour.
Wonder Create goes further, allowing aspiring restaurateurs to use AI to invent a restaurant concept, generate a menu and have Wonder’s kitchens prepare and deliver the food. Sadly, if you’ve dreamed of creating a “high-energy Mexican-Greek street-food bowl concept,” that’s already Wonder’s own example.
Fully committing to the bit, Founder Marc Lore has surrendered about 90% of his diet to AI and no longer has to worry about whether he should have four, five or six raw walnuts to optimize his breakfast. (The answer is five.) If such foodmaxxing stresses you out, at least it’s not Huel.
Anyway, let’s note the aforementioned AI agent order statistics for the record, as they might become one of those quaint technological milestones we’ll look back on in 2036, much as we do the guy who bought two Papa John’s pizzas for 10,000 Bitcoin in 2010—worth some USD 750 million at today’s rates.
