The 'Frictionless' Fallacy: Why AI Agents Are Failing the DoorDash Test

In the race to automate every click of the consumer journey, the industry has chased a 'frictionless' ideal that often ignores the reality of the user experience. As Ars Technica first reported, this is increasingly evident in the current failure of third-party AI agents to gain traction in the food delivery space.

**Opinion:** From a commerce operator's perspective, the lack of adoption here isn't a failure of the technology, but a failure of the UX logic. Ordering food is a high-stakes, high-nuance activity. When a customer is hungry, the 'friction' of opening a dedicated app is negligible compared to the risk of an AI agent misinterpreting a craving or missing a critical order detail. True efficiency isn't removing the human from the loop; it's optimizing the loop for the specific emotional state of the buyer.

According to reporting from Ars Technica, the numbers back this up. During a third-quarter earnings call on Thursday, DoorDash co-founder and CEO Tony Xu noted that third-party AI agents have generated only "several hundred orders" since they became operational. To put that in perspective, DoorDash handled nearly 1 billion total orders in the third quarter of 2026 alone. Xu characterized this as a "very low use case."

While DoorDash recently introduced a Model Context Protocol (MCP) connector to provide a standardized way for third-party AI tools to place orders, the feature is currently limited to a small group of "early beta partners" and is primarily utilized for corporate lunch orders. Without this connector, agents must navigate the platform via web interfaces, adding a layer of complexity that most consumers are unwilling to endure for a simple meal.

Interestingly, Xu suggested that the real value of AI agents lies not in these quick transactions, but in "long consideration purchases" and major decisions where an agent can proactively work in the background. For the immediate need of food, Xu noted that customers tend to go directly to DoorDash.

However, the data suggests that AI *can* drive growth when integrated directly into the platform's UX rather than acting as a third-party intermediary. Ars Technica reports that DoorDash's internal "Ask DoorDash" tool—which allows users to build carts from recipe photos or vague descriptions—has seen significant success. The company claims these baskets are built "five times faster" and boast nearly 50% higher basket values and 60% more unique items than traditional orders. Furthermore, nearly half of these AI-assisted orders are placed at local restaurants the consumer has never previously tried.

Despite the potential for future growth, Xu remains pragmatic about the current state of agentic commerce, stating that while one should not "fight gravity," one also shouldn't be "delusional."