Tony Xu says AI-native rivals do not scare him because DoorDash coordinates an app, logistics and physical operations that software alone cannot replace.
DoorDash plans to partner with foundation model companies to digitize physical-world data like parking spaces and Safeway inventory.
Xu still does 15 minutes of customer support daily, and DoorDash's 'one team, one fight' culture puts customer wins above job descriptions.
DoorDash CEO Bets on Physical World Over AI Agents
DoorDash is a physical-world logistics company. Tony Xu says it carries about a million of the tens of millions of items people buy, so growth depends on digitizing everything inside a city and bringing it to the customer. That requires data not online, like parking spaces and Coke cans on a shelf, and an end-to-end system of app and logistics that a digital-only AI competitor cannot replicate. Xu frames the edge as customer obsession rather than competitive fear.
Takeaways
DoorDash serves only a sliver of demand. Xu says DoorDash carries about a million of the tens of millions of items people buy, serving roughly 5% of expressed needs. The goal is to bring everything inside a city to the customer, which he frames as growing the GDP of those cities.
AI agents still lack human judgment. Xu says today's tools are the worst version of what is coming, but agents are not yet suited to workflows needing human judgment or with no verifiable answer. The key is the end-to-end job for the customer.
DoorDash must digitize offline physical data. Data like Coke cans for sale at the Safeway on 4th Street and the cheapest size 10.5 shoes for a 5-kilometre race does not exist online. Xu says DoorDash will probably team with foundation model companies to build models from it.
AI competitors cannot replace physical operations. Asked whether an AI-native DoorDash scares him, Xu says the company is an app and a logistics algorithm coordinated with customer service. A digital solution cannot replace an end-to-end system with physical parts, and he evaluates new technology against where customers are most vulnerable.
Customer obsession shows where DoorDash is weak. Aditya Agarwal notes Xu is customer-obsessed rather than competitively focused. Xu says if he looked at every metric from a customer's perspective instead of the output metrics reported to Wall Street, he would quickly know where DoorDash is most vulnerable.
Doing the work teaches what algorithms miss. Xu still does about fifteen minutes of customer support a day and deliveries every year. Doing deliveries himself taught him that assigning the closest driver to a restaurant may not be best if the food is not ready. Four founders did the same.
DoorDash's operating playbook
Bias for action means settling debates by shipping something.
One team, one fight means your job is to help the customer win, even if that means doing deliveries or customer support.
Every year the company asks what it would do differently if starting from scratch.
Xu deletes his own account to test the new-user flow.
Xu reads thousand-word emails from Dashers.
The non-consensus years
Three years without raising capital.
Rejected by more than 100 investors.
Lost 20 to 25 percent of the team.
Metrics convinced 75 to 80 percent to stay.
Tap a name to go there · tap a line to see why they are linked