DoorDash said it’s only seen ‘several hundreds’ of orders from AI agents

DoorDash said it has seen only “several hundred” orders from third-party AI agents, compared with an average of 3 million deliveries per day reported in 2025. CEO Tony Xu said customers still predominantly order through DoorDash’s app, while the company is exploring agent-based ordering and potential advertising within third-party agents.

Analysis

The key economic question is not whether an agent can submit an order, but who owns discovery, customer data, and the economics of the transaction. If agents become a meaningful entry point, DoorDash could face weaker control over repeat engagement and cross-selling; conversely, its merchant access and delivery execution remain valuable even if the interface shifts. Advertising inside agents is a possible offset, not yet evidence of a monetizable channel. The company’s own ordering tools establish strategic optionality, but launches alone do not demonstrate adoption or attractive unit economics.

Amazon’s restrictive posture protects its customer interface and first-party data, but risks ceding incremental demand if consumers increasingly delegate shopping. Walmart’s openness could acquire customers at the margin, while also exposing its assortment and pricing to easier comparison. Neither posture is proven superior: outcomes depend on whether agents deliver incremental orders or merely reroute existing ones, and on the fees, data rights, and service standards platforms can negotiate.

Near term, the disclosed activity is too small to justify changing earnings assumptions or paying an AI-agent premium; treat this as a distribution experiment, not a demand inflection. Over 1–3 months, watch for reported order mix, repeat use, agent access restrictions, and any disclosed economics. Over 6–18 months, the structural risk is interface disintermediation, but fulfillment reliability and merchant relationships may limit platform substitution. The thesis changes if agent-originated orders scale materially while DASH reports weaker app engagement, higher acquisition costs, or lower contribution economics. Current evidence supports no event-driven position.

Key Decisions for Investors

• Do not initiate a DASH position on agent-ordering headlines alone; the evidence does not yet support an earnings or valuation revision.
• Track DASH disclosures for agent-sourced order share, repeat rates, customer acquisition costs, and whether third-party access carries fees or data-sharing conditions. Rising volume without profitable repeat use would not validate the opportunity.
• Monitor AMZN and WMT for changes in agent access, conversion, and customer acquisition—not simply whether they permit agents. A reversal in access policy or evidence of material incremental demand would be a more actionable catalyst.
• Falsify the current immateriality view if agent-originated orders scale rapidly and DASH reports measurable monetization or retention benefits; reassess downside if app engagement weakens or agent intermediation pressures order economics.