News Wikipedia's Operator Tells Court AI Training Should Stay Fair Use - but Chatbots That Replace the Original Should Pay AI Agents DoorDash Has Seen Maybe Several Hundred Orders Total From Third-Party AI Agents - Against 3 Million Deliveries a Day Business DeepSeek Closes a Second Round of About 105.5 Billion Yuan, Nearly $14.8 Billion, at a Roughly $86 Billion Valuation Security A Backdoored Claude Code Extension Sat in Cursor's Marketplace for 11 Hours - and Uninstalling It Does Not Remove It Voice & Audio Spotify Wants $5 More a Month to Let Fans Remix Its Catalog With AI, Starting Around the Grammys AI Agents Coursera's First Employee and a TensorFlow Co-Creator Raise $11 Million to Give Every AI Agent an ID Badge News UK Regulator Opens Probe Into Meta's AI Content Terms as Publishers Fight on Two Fronts Business A Data Center Every 48 Hours: Runware Wants to Build AI Capacity on a Car Factory Line News Wikipedia's Operator Tells Court AI Training Should Stay Fair Use - but Chatbots That Replace the Original Should Pay AI Agents DoorDash Has Seen Maybe Several Hundred Orders Total From Third-Party AI Agents - Against 3 Million Deliveries a Day Business DeepSeek Closes a Second Round of About 105.5 Billion Yuan, Nearly $14.8 Billion, at a Roughly $86 Billion Valuation Security A Backdoored Claude Code Extension Sat in Cursor's Marketplace for 11 Hours - and Uninstalling It Does Not Remove It Voice & Audio Spotify Wants $5 More a Month to Let Fans Remix Its Catalog With AI, Starting Around the Grammys AI Agents Coursera's First Employee and a TensorFlow Co-Creator Raise $11 Million to Give Every AI Agent an ID Badge News UK Regulator Opens Probe Into Meta's AI Content Terms as Publishers Fight on Two Fronts Business A Data Center Every 48 Hours: Runware Wants to Build AI Capacity on a Car Factory Line

DoorDash Has Seen Maybe Several Hundred Orders Total From Third-Party AI Agents - Against 3 Million Deliveries a Day

DoorDash Has Seen Maybe Several Hundred Orders Total From Third-Party AI Agents - Against 3 Million Deliveries a Day

On its Q3 2026 earnings call, DoorDash CEO Tony Xu said third-party AI agents including Meta's Muse, Instinct and Fo have generated "maybe several hundred orders" in total - a rounding error against the 3 million deliveries the platform completes daily. Xu said users still order predominantly through DoorDash's own apps, while the company builds its own agent stack: Ask DoorDash, a CLI agent, a text-to-order agent and an MCP ordering connector.

The agentic commerce narrative has been running hot all year: AI assistants will browse, decide and order on behalf of users, rewiring how consumers transact online. DoorDash, one of the largest delivery platforms in the United States, has now put a hard number on how much of that future has actually arrived - and it is tiny.

On the company's third-quarter earnings call, CEO Tony Xu was asked about ordering through third-party AI agents. "This is something that is still a very dynamic area. And my take on all things technology is you never want to fight gravity. But at the same time, you don't want to be delusional," Xu said. "What I mean by that is if you look at where AI agents are today, we're seeing maybe several hundred orders in total from some of these third-party agents."

The scale mismatch is stark. In its 2025 report, DoorDash said it completes an average of 3 million deliveries per day. Even granting generous assumptions about growth since then, orders arriving through external agents - Xu named Meta's Muse, Instinct and Fo - amount to a rounding error, several hundred in total across months of partnerships and integration announcements.

Xu added that users still place orders predominantly through DoorDash's own app, and that the company wants to keep learning about agentic use cases while building "a healthy relationship" with third-party agent makers. He also flagged a future business angle: DoorDash could eventually advertise inside these agents, the way it advertises across other demand channels today.

The company is not sitting out the shift - it has simply chosen to build its own on-ramps. DoorDash launched the Ask DoorDash chatbot in June, letting users order through prompts and photos; a command-line agent in July, letting developers order food from the terminal; and a text-to-order agent in September. It has also released an MCP connector that enables ordering from corporate chatbots for enterprise catering. The pipes for agent-mediated ordering exist on both sides; what is missing, so far, is demand flowing through them.

Platforms are splitting on how to treat outside agents. Amazon has been blocking personal shopping agents from its storefront, while Walmart rolled out the welcome mat in late October, hoping to pull business from competitors by being the easiest place for an agent to shop. DoorDash's disclosure suggests the blocking-versus-welcoming debate may be outrunning reality: there is not yet enough agent traffic for either posture to matter much commercially.

The takeaway for the industry is a dose of arithmetic. Integration announcements, protocols and partnerships have multiplied faster than orders. Xu's framing - respect the trend, do not hallucinate its current size - may become the template for how consumer platforms talk about agents in 2027 earnings calls: as a strategic option worth building for, and a revenue line still waiting to exist.

Comments (0)

Log in to join the discussion

Log In

No comments yet