DoorDash said it’s only received ‘several hundred’ orders from AI agents

DoorDash said it has seen only “several hundred” orders through third-party AI agents, compared with an average of 3 million deliveries daily in its 2025 report. CEO Tony Xu described agent-based ordering as a dynamic, early-stage area; DoorDash is exploring relationships with third-party agents and potential advertising, while also developing its own ordering agents.

Analysis

The near-term read-through is limited: agent-mediated ordering is too small to establish a demand or earnings trend. The strategic issue is control of discovery, customer data, and checkout—not order volume today. If agents become the front end, DoorDash could lose app engagement and advertising attribution even if it still fulfills the delivery. Its own ordering tools and corporate MCP are therefore defensive options to preserve access, but the company’s stated future advertising opportunity is not yet evidence of incremental monetization.

Walmart’s openness may help it capture incremental orders or steal share, but it also risks ceding customer intent and comparison shopping to third-party interfaces; more orders would not automatically mean better economics. Amazon’s access restrictions protect its owned customer journey for now, though they could become a disadvantage if consumers adopt agents that route across retailers. These are conditional strategic exposures, not demonstrated financial impacts.

Over the next 1–3 months, watch for disclosed agent order volumes, integration terms, and evidence of incremental versus shifted orders. Over 6–18 months, the key question is whether platforms can charge for agent-originated demand or whether agents commoditize fulfillment and intensify price competition. The headline-scale concern looks premature; the risk is underappreciated only if adoption accelerates and agents control checkout. No directional trade is justified on this evidence alone.

Key Decisions for Investors

• No immediate trade in DASH, WMT, or AMZN: the disclosed activity is not yet a material earnings signal, and the eventual economics of agent referrals remain unverified.
• Set a DASH watch item: reassess if management reports sustained agent-originated order growth, material app-engagement changes, or specific referral/advertising economics. Falsifier for the disintermediation concern: agent use grows while DASH retains customer relationships and monetizes those orders without weakening unit economics.
• Track WMT integrations for incremental customer acquisition versus order migration from Walmart’s own channels; openness is not inherently positive if third-party agents capture discovery and bargaining power.
• Revisit the sector view if major platforms change access policies or disclose agent-order volumes and economics. A shift toward cross-retailer agent checkout would raise longer-term customer-acquisition and price-transparency risk for delivery and retail platforms.