Building an Autonomous Delivery Experience with DoorDash Co-Founders Andy Fang and Stanley Tang
DoorDash co-founders Andy Fang and Stanley Tang discuss their expansive vision for the company beyond food delivery, including agentic commerce features that enable natural language ordering, autonomous delivery robots (DOT), and a multimodal fleet strategy. They emphasize that their deep operational knowledge and 10 billion deliveries of data provide unique advantages in building AI and robotics solutions for real-world physical commerce.
Summary
Andy Fang and Stanley Tang, co-founders of DoorDash, discuss the company's evolution into a multi-modal logistics and commerce platform powered by AI and robotics. On the AI front, they launched Ask DoorDash, an agentic commerce feature enabling natural language queries for food and groceries. Initial data shows 50% of restaurant orders through this feature are from new merchants, and grocery basket sizes are 40% larger, indicating latent demand for easier discovery and interaction. The founders attribute this to people's preference for natural conversation over keyword search optimization.
They detail DoorDash's long-term robotics strategy, which began in 2018 as experimentation. Initial partnerships with various autonomy companies (sidewalk robots and robo-taxis) revealed that neither form factor matched DoorDash's use case: suburban deliveries averaging 3-5 miles completed in 15 minutes. This led them to build DOT in-house—a 300-pound, 20 mph autonomous vehicle designed specifically for delivery in bike lanes and roads, operating for nearly two years in Phoenix and achieving full Level 4 autonomy last year.
The founders emphasize that successful robotics and AI require deep understanding of use cases rather than technology-first approaches. They stress the critical importance of real-world data and operational complexity: three billion annual DoorDash deliveries are all different, spanning diverse geographies, merchant types, and delivery contexts. This complexity revealed through live operations exposed unforeseen challenges—from torque distribution when wheels rest on leaves to GPS pin inaccuracy at apartment complexes to battery management during extreme braking.
On scaling challenges, they identify three components: autonomous capability (increasingly solvable), operations (adapting to different merchant interfaces and fleet management across cities), and hardware (manufacturing and supply chain reliability). They partnered with Alize for manufacturing expertise.
Regarding AI's impact on their workforce, they've acquired Metis to infuse AI-native thinking across the 10,000+ person company. They track AI spending (which increased 20x from January to June) through benchmarks like DashBench for coding tasks, and are exploring ROI measurement across finance, accounting, and analytics functions. They note that enterprise data often reveals model performance gaps not seen in cleaned datasets.
Fang emphasizes that the long-term vision isn't replacing dashers with robots, but rather creating a multimodal fleet to meet growing demand. With 9 million dashers and 25% YoY growth, they predict needing even more delivery capacity in 10 years despite robotics expansion. They expect delivery demand will increase due to affordability gains, and they're exploring agent-first experiences like the DoorDash CLI, including automated pantry restocking based on camera monitoring.
About this episode
DoorDash is not just a delivery company. From its inception, co-founders Andy Fang and Stanley Tang operated it as a robotics and autonomy company. Andy and Stanley join Sarah Guo to explain how autonomous tech and AI are reshaping consumer habits, commerce, and delivery. Andy and Stanley talk about the rollout of Ask DoorDash, a natural-language interface that’s driving both restaurant discovery and larger grocery orders. They also discuss Dot, their in-house autonomous delivery robot that has operated in Phoenix for over two years, and how it highlights the operational and hardware challenges they have faced and solved in autonomous tech. Andy and Stanley also speak about the “first and last 100 feet problem” in autonomous delivery, why multimodal strategies are the key to success, scaling autonomy and operations, and why they believe that more Dashers, not fewer, are the future of DoorDash. Sign up for new podcasts every week. Email feedback to [email protected] Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @stanleytang | @andyfang | @DoorDash Chapters: 00:00 – Andy Fang and Stanley Tang Introduction 00:34 – Agentic Commerce and Behavioral Changes 03:52 – Next Steps for Ask DoorDash 06:54 – Investing in Robotics and Autonomy 16:31 – Building Autonomous Tech in the Physical World 21:20 – Dot: DoorDash’s Autonomous Delivery Robot 22:08 – Collecting Realistic Data 25:48 – Why Work at DoorDash 28:04 – Challenges in Scaling Up Autonomy 39:30 – Productivity Benchmarks 44:56 – Future of Agentic Commerce 49:10 – Conclusion
Key Insights
- DoorDash found that 50% of Ask DoorDash restaurant orders come from merchants customers had never ordered from before, revealing significant latent demand for restaurant discovery that wasn't being served by the traditional interface.
- The founders determined that neither existing sidewalk robots (too slow at 2-3 mph) nor robo-taxis (designed for people, not packages, and too heavy at 4,000 lbs) matched their use case, forcing them to build DOT in-house as a 300-pound, 20 mph vehicle specifically designed for suburban delivery.
- Real-world robotics operations exposed unforeseen physical challenges not discoverable through simulation—such as unequal torque distribution when some wheels rest on leaves while others stay on asphalt, and incorrect battery regen braking during emergency stops.
- DoorDash's competitive advantage in robotics deployment stems from possessing the only dataset containing actual historical drop-off locations and merchant pickup points from 10 billion prior deliveries—data that doesn't exist elsewhere, including Google Maps.
- The founders argue that the hardest problems in scaling autonomy are no longer autonomy itself, but rather operational adaptation to different merchant types and geographies, and hardware manufacturing and supply chain reliability.
- Andy Fang reported that DoorDash's AI spending flattened after increasing 20x from January to June 2024, and they identified opportunities to use cheaper open-weight models for lower-value tasks rather than always using closed-weight models.
- Stanley Tang predicts that despite robotics expansion, DoorDash will have more human dashers in 10 years, not fewer, because the business is growing 25% YoY and will need additional delivery capacity from all modalities combined to meet demand.
- The founders describe the founding philosophy as experiment-driven rather than technology-first, beginning robotics efforts in 2018 as a 'Skunk Works project' with minimal resources, only scaling up after learning through partnerships and real-world deployments revealed what was actually needed.
Topics
Transcript
Hi listeners, welcome back to No Priors. Today I'm here with Andy Fang and Stanley Ting, co-founders at DoorDash. We talk about how you can ask DoorDash in natural language for food and groceries, what that means for the future of agentic commerce, their delivery robot DOT, how DoorDash has been a robotics company for the last eight years, the data advantages of their network, and what all this means for 9 million dashers and 3 billion deliveries a year. Welcome. Andy, Stanley, thank you so much for being here. Real excited to talk to you about all the crazy stuff DoorDash is doing. I thought we could start with what's going on with agentic commerce at DoorDash. I feel…
Full transcript available for MurmurCast members
Sign Up to AccessMore from No Priors: Artificial Intelligence | Technology | Startups
Re-Founding Incumbents for the AI Era with Sequence Holdings Co-Founder and CEO Michael Lee
Michael Lee, co-founder and CEO of Sequence Holdings, discusses his permanent holding company model for acquiring and refounding incumbent businesses using frontier AI and engineering talent. He shares lessons from Sequence's first investment in BankSouth and their recent $7.7 billion take-private of insurance broker Baldwin, explaining why this model differs fundamentally from traditional private equity and venture approaches.
Why Diffusion Will Win AI Inference with Inception Co-Founder and CEO Stefano Ermon
Stefano Ermon, Stanford professor and CEO of Inception, discusses why diffusion models will dominate AI inference, arguing they offer fundamental advantages over autoregressive models for parallelization and speed. His company is scaling diffusion-based language models as a competitive alternative to frontier labs, with early commercial deployments showing 10x faster inference while maintaining comparable quality.
Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Brian Armstrong discusses Coinbase's evolution into an "everything exchange" supporting agentic finance, stablecoins, and tokenization, while also exploring his work at New Limit on cellular reprogramming for longevity. He argues that AI agents will need their own financial accounts and that crypto rails are essential for enabling micro-transactions that traditional payment systems cannot support economically.
Redefining Chip Architecture with Arm CEO Rene Haas
Arm CEO Rene Haas discusses Arm's evolution from IP licensing to building physical chips, the critical role of CPUs in AI infrastructure, and how AI tools are accelerating chip design cycles. He addresses supply chain constraints, robotics opportunities, export controls, and why data center development is essential for US technological leadership.
Rethinking Legacy Data Infrastructure with Eon Co-Founders Ofir Ehrlich and Gonen Stein
Ofir Ehrlich and Gonen Stein, co-founders of Eon, discuss how AI and agents are fundamentally transforming enterprise data infrastructure. They explain that data has become the most valuable asset for organizations, and new tools are needed to securely map, classify, and activate this data for AI applications while managing risks from autonomous agents with legitimate system access.