DoorDash is betting that the future of food delivery involves typing a single sentence — but its own numbers suggest that future is arriving slowly. The company this week unveiled a text-to-order AI agent that lives inside Apple Messages, allowing customers to fire off a casual request like "order my usual" and let a large language model handle the rest. On the same day, executives disclosed on the company's third-quarter earnings call just how few people are actually letting AI agents shop for them.
A few hundred orders against nearly a billion
Asked by an analyst about agentic commerce, co-founder and CEO Tony Xu offered a figure that was striking for its modesty: AI agents have generated "several hundred orders" on DoorDash since the capability went live. DoorDash handled just shy of 1 billion orders in the third quarter alone.
"AI agents are generating several hundred orders across the time that they've been up and running." — Tony Xu, DoorDash CEO
Xu called it a "very low use case," an understatement that Ars Technica noted was doing considerable heavy lifting. His explanation doubles as a strategic thesis: agents shine at "long consideration purchases" and major life decisions, where an assistant can proactively work in the background on your behalf. Ordering a burrito is not that.
"When it comes to a task that's as simple as ordering food, we tend to see that come directly to DoorDash." — Tony Xu
That tension sits at the heart of the announcement. DoorDash is simultaneously telling investors that agent-driven ordering is a rounding error and shipping new agent products designed to make it more common.
What the new agent actually does
The headline feature, reported across The Verge, The Next Web, ExtremeTech, MacObserver and others, is a conversational ordering agent that runs inside Apple Messages. Rather than opening the DoorDash app, browsing menus and assembling a cart, a user can text a natural-language instruction — the company's showcase example being "order my usual" — and the agent infers intent, pulls from order history, confirms the details and places the order.
It is, in effect, an attempt to move DoorDash from an app you open to a service you message, slotting into the same thread where you talk to friends, family and colleagues. That placement is deliberate: messaging apps are where users already spend their attention, and reducing friction is the oldest growth lever in delivery.
Office ordering and a drone roadmap
The iMessage agent was not the only reveal. Reports from Rutland Herald and Seeking Alpha indicate DoorDash is also pushing group and office ordering into the AI tools that teams already use at work, extending the same conversational logic from individuals to workplace lunch runs. Seeking Alpha additionally reported that DoorDash used the moment to broaden its drone delivery plans — a reminder that the company's long-term cost story depends as much on logistics automation as on conversational interfaces.
Meta's $199.99 question
Coverage from MSN added a wrinkle that reframes the entire debate: Meta reportedly wants $199.99 a month for an AI agent capable of ordering your DoorDash. That price point — roughly the cost of a streaming bundle, a gym membership and a few burritos — amounts to a bet that consumers will pay handsomely for delegation itself.
If that bet is wrong, the ceiling on agentic commerce may be structurally low. If it is right, DoorDash's "several hundred orders" becomes a seed rather than a punchline. Either way, it highlights a divergence in how the industry is pricing agentic convenience: DoorDash is giving the capability away inside its own product, while Meta is attempting to sell it as a premium subscription.
Xu's counterintuitive labor argument
Perhaps the most notable comment from the earnings call, flagged by The Next Web, was Xu's claim that AI agents will ultimately increase demand for human labor. The logic runs roughly as follows: if agents lower the friction of ordering, they expand the total volume of commerce, which requires more dashers, more restaurant staff and more logistics workers to fulfill it. It is an optimistic reading of automation — one that positions AI as a demand-side accelerant rather than a labor-substitution engine.
Skeptics will note that the same argument has accompanied nearly every automation wave in the past decade, and that delivery economics are driven heavily by cost-per-order, which automation of any kind is designed to reduce.
How the press framed it
- Enthusiast framing: The Verge, ExtremeTech and MacObserver focused on the novelty and convenience of texting "order my usual," treating the agent as a consumer-friendly upgrade.
- Skeptical framing: Ars Technica led with the adoption data, contrasting DoorDash's grand agentic ambitions against a usage figure measured in the hundreds.
- Snark framing: IGN's headline — that ordering DoorDash is apparently "too hard," so an AI will do it for you — captured a broader consumer fatigue with AI being bolted onto tasks that were never broken.
- Business framing: Seeking Alpha and The Next Web emphasized strategy: drone delivery, labor demand and the long game of agentic commerce.
The larger context
DoorDash's dilemma is the industry's dilemma. Every major platform is racing to expose its catalog to AI agents — through Apple Messages, through workplace chat tools, through third-party assistants — because the fear of being disintermediated by an LLM is greater than the fear of building for a market that does not yet exist. But the data so far suggests that consumers, given a choice, still prefer the app. Convenience, it turns out, is not the same as frictionless.
The company's wager is that habits shift slowly and then all at once. The text-to-order agent launches into a market where the behavior is, for now, close to nonexistent. Whether that changes depends less on how clever the model is and more on whether messaging a bot ever feels easier than tapping a burrito icon.
