One of the earliest debates we had at DoorDash was whether we ought to first build the consumer app or the Dasher app, the product that Dashers use to make deliveries. It seemed so intuitive that we ought to first build the consumer app because isn't that what we are? Yet, as founders fulfilling deliveries everyday at that time, we knew that bringing you a burrito — on time, in the condition you'd want, every single time — was a lot harder than it sounds.

Systems are orchestration of products whose job is to provide a complete and high quality experience for its customers. When done right, building cohesive systems allows us to build the best customer experience at the lowest cost. To build great systems, you have to operate at the lowest level of detail. This is why, since our founding, we've asked employees to WeDash, a tradition in which we fulfill deliveries ourselves to understand how our products interact and how our systems succeed or fail.

While AI agents haven't yet achieved their full potential, it's clear that adding the reasoning power behind LLMs can produce more intelligent systems. But this will require careful investigation as LLMs require verifiable responses to produce great outcomes, and things can go haywire when they don't have the right context or aren't connected to the appropriate workflows. AVs by land and air will undoubtedly become a part of our systems, and we are pushing aggressively to make this happen — including DoorDash Dot, our purpose-built vehicle for delivery. Regardless of technology, AVs, AI agents, or chatbots alone cannot solve the entirety of a customer's problem. These products will become part of our systems at DoorDash.