Building an Autonomous Delivery Experience with DoorDash Co-Founders Andy Fang and Stanley Tang

What It Covers

DoorDash co-founders Andy Fang and Stanley Tang detail how the company is deploying conversational AI commerce and its in-house autonomous delivery robot, Dot, across a network processing 3 billion deliveries annually, revealing data advantages, multimodal fleet strategy, and the operational complexity of scaling physical-world robotics.

Key Questions Answered

• •Conversational commerce conversion: Ask DoorDash drives 50% of restaurant-search sessions toward orders from restaurants users have never previously ordered from — historically one of DoorDash's hardest metrics to move. On the grocery side, natural-language interactions produce basket sizes roughly 40% larger than traditional tap-based browsing, driven by use cases like fridge-photo meal planning and dietary-constraint filtering.
• •Autonomous delivery form factor: DoorDash's Dot robot weighs 300 pounds, travels up to 20 miles per hour, and operates on roads, bike lanes, and sidewalks — a profile deliberately between slow sidewalk robots (2–3 mph, range-limited) and 4,000-pound robotaxis. The three-to-five-mile suburban delivery radius in dense markets like Phoenix defined the specification before any hardware was built.
• •Use-case-first robotics strategy: DoorDash spent several years partnering with external autonomy companies before building Dot in-house, concluding that most robotics startups build technology first and retrofit a problem later. The productive inversion — define the delivery use case, work backwards to hardware — is the same customer-first methodology DoorDash applied from its founding as paloaltodelivery.com with eight PDF menus.
• •Proprietary last-100-feet data: GPS pins for delivery addresses are unreliable, especially at apartment complexes. DoorDash's historical Dasher drop-off location data — unavailable in Google Maps or any external source — trains Dot to navigate the precise driveway, gate, or entrance for each address, making this dataset a structural moat for autonomous last-mile delivery at scale.
• •Multimodal fleet economics: Rather than replacing Dashers, DoorDash projects more human Dashers in ten years, not fewer, because 25% annual business growth across 9 million current Dashers cannot be met by humans alone. The strategy assigns delivery modality by use case: Dot handles three-to-five-mile suburban runs, drones cover lightweight rural orders, and human Dashers manage complex multi-item grocery pick-and-pack tasks.