From Chatbots To Checkout: How Much Are AI Agents Shopping?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: From Chatbots To Checkout: How Much Are AI Agents Shopping? on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

DoorDash CEO Tony Xu said third-party AI agents had generated “maybe several hundred orders in total” since they began operating on the platform. DoorDash handled 993 million orders in the third quarter of 2026, while Amazon and Visa executives also described agent-led shopping as limited. Meta has reported stronger shopping activity for its Muse agent, but its figures are company-reported and not directly comparable.

DoorDash CEO Tony Xu said third-party AI agents had generated “maybe several hundred orders in total” since they began placing orders on the platform, a small tally beside the 993 million orders DoorDash recorded in the third quarter of 2026. His comments, made on the company’s October 29 earnings call, offer a concrete measure of how limited agent-led ordering remains at a major delivery service, despite growing attention from technology and retail companies.

Xu gave the figure while responding to analyst questions about whether DoorDash would build its own AI agents or work with general-purpose tools such as Meta’s Muse. He described agent shopping as a “dynamic area”, but said that for a straightforward task such as ordering food, most customers still use the DoorDash app. Later in the call, he repeated that agents had generated several hundred orders across the time they had been available and called it a “fairly low use case.”

DoorDash reported 993 million total orders in the third quarter, up 28% from a year earlier. That total covers the quarter; Xu’s rounded estimate covers the full period that third-party agents have been operating, so the figures do not share the same time window. If “several hundred” is treated as 1,000 and all those orders are placed within the quarter, the comparison would be roughly one agent order per million total orders. That is an illustrative calculation, not a company-reported rate.

Executives at other companies gave similarly cautious accounts that week. Amazon CEO Andy Jassy said few customers were doing much shopping through outside agents, describing the experience as neither good nor more convenient than shopping directly. Visa CEO Ryan McInerney referred to an “albeit limited number” of agent transactions and said trust remained a barrier to wider use. Meta CEO Mark Zuckerberg, by contrast, said shopping was the most common use of its Muse agent. These statements describe different services and measures, and should not be read as a single industry-wide tally.

At a glance
reportWhen: Earnings-call comments made October 29,…
The developmentDoorDash’s CEO disclosed on the company’s October 29 earnings call that third-party AI agents had placed several hundred orders in total, highlighting the gap between agent-shopping claims and current use at major services.
Thorsten Meyer AI
DoorDash · Q3 2026

Infographic 01 · The number

993 million orders. “Several hundred” from AI agents.

What DoorDash’s CEO said about orders placed by outside AI agents, next to the company’s own order count.

993Morders in Q3 2026, up 28% on a year earlier (DoorDash)
≈ 10.8Morders per day (my calculation, 92 days)
100s“several hundred” orders from outside AI agents, in total since they started (CEO)
≈ 1 in 1Meven at 1,000 agent orders, all in one quarter (my calculation)

Drawn to scale

Q3 2026 orders vs. a generous upper bound for agent orders

  • All ordersDoorDash, Q3 2026993M
  • Via outside AI agentsupper bound: 1,000, any period≤ 1,000

The second bar is not missing. At this scale it is about one ten-thousandth of one percent wide.

“You never want to fight gravity. But at the same time, you don’t want to be delusional … we’re seeing maybe several hundred orders in total from some of these third-party agents.”Tony Xu, CEO of DoorDash · Q3 2026 earnings call, 29 Oct 2026

Sources: DoorDash Q3 2026 results (29 Oct 2026); DoorDash earnings call (29 Oct). Ratio and per-day figure: my calculation. “Several hundred” is a spoken, rounded figure.
Powered by Thorsten Meyer AI

Agent Orders Lag Platform Scale

The comments put a scale check on predictions that AI agents are about to redirect substantial shopping activity. At DoorDash, the disclosed volume is tiny relative to the company’s ordinary ordering business, suggesting that the existence of agent tools has not yet translated into widespread use for food delivery. For customers, the immediate implication is that the familiar app remains the dominant route to ordering, at least on this platform.

The figures also matter to retailers and advertisers. If customers make purchases inside an agent interface rather than a retailer’s own app or site, the retailer may lose some influence over product discovery, checkout, customer data and advertising. Xu said DoorDash sees potential advertising opportunities within AI agents, but added that current volumes do not yet create that value. The near-term commercial question is not only whether agents can complete purchases, but whether enough people use them to change how companies earn revenue.

Low order counts do not settle what the market will look like later. They do, however, distinguish present behavior from company plans and promotional claims. A service may attract sign-ups or shopping queries without producing a comparable number of completed purchases. For businesses evaluating investments, completed transactions and repeat use are more direct indicators of adoption than attention alone.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

Different Agents, Different Measures

The discussion has intensified as companies add shopping functions to general-purpose assistants and build tools intended to interact with retailers. The source material says Meta reported that more than 10 million people in the United States and Canada had set up Muse in about six weeks. Zuckerberg said shopping was its most common use and that more than a quarter of users engaged in it, with an average purchase size above $200. Those are Meta’s own figures; the company did not provide a total sales value in the cited remarks.

Those Muse figures are not directly comparable with DoorDash’s order count. Meta described user activity and average purchase size, while Xu discussed completed orders from third-party agents on DoorDash. The available account does not specify a shared measurement period or a common definition of an agent-led purchase. Visa’s reference to agent transactions and Amazon’s comments about customers shopping through outside agents use still other measures.

DoorDash has also described a broader strategy that includes building its own AI tools while investing in the logistics and physical infrastructure behind delivery. The source material characterizes that approach as a focus on “atoms vs. bits” and cites warehouses, autonomous delivery and a proposed $17 billion deal to buy Instacart. Those strategic moves do not establish that agent ordering is already widespread; they show that companies are preparing for changes in how orders may be found and fulfilled.

““We’re seeing maybe several hundred orders in total from some of these third-party agents.””

— Tony Xu, DoorDash CEO, on the October 29, 2026 earnings call

The Count Has Important Limits

Xu did not give an exact count, a start date for the measurement, or a breakdown by agent, order type or customer. “Several hundred” is a rounded estimate, and he said it covered the time the agents had been operating rather than a defined quarter. The source material does not establish how many attempts failed, whether repeat orders were included, or how DoorDash classifies an order as agent-generated.

Meta’s reported Muse activity also leaves open how many shopping interactions became completed purchases and how its figures compare with transactions on DoorDash or other services. The companies’ descriptions are not a common industry dataset, and the source provides no independent audit of Meta’s figures. It remains unclear whether low use reflects limited customer interest, friction in the checkout process, restricted access to retailer systems, concerns about payment and trust, or some combination.

Watch for Orders, Not Sign-Ups

Future earnings calls and company disclosures may show whether agent-led orders grow from the low levels Xu described. Useful indicators would include a defined reporting period, exact transaction counts, the share of completed checkouts, repeat purchasing and clearer separation between shopping activity and purchases. Without those details, comparisons between platforms will remain difficult.

DoorDash’s comments also leave open whether the company will enable more third-party agents, develop its own ordering tools, or both. Xu said advertising opportunities could emerge inside agents if volumes grow, but did not describe a current revenue-sharing arrangement or provide a timetable. For now, the confirmed development is a small reported order count—not a forecast of how quickly it may change.

Key Questions

How many orders have AI agents placed through DoorDash?

CEO Tony Xu said third-party agents had generated “maybe several hundred orders in total” across the time they had been operating. He did not give an exact number or a specific start date.

How does that compare with DoorDash’s total orders?

DoorDash recorded 993 million orders in the third quarter of 2026. Because Xu’s estimate covers a different, unspecified period, a direct rate cannot be calculated from the disclosed figures. Treating “several hundred” as 1,000 and placing all of those orders in the quarter gives an illustrative comparison of about one per million total orders.

Does Meta report more shopping through its Muse agent?

Meta CEO Mark Zuckerberg said shopping was Muse’s most common use and that more than a quarter of users engaged in shopping, with an average purchase size above $200. These are Meta-reported figures, and the source material does not give total sales or an independent audit. They measure activity differently from DoorDash’s agent-order count.

Why might customers not be ordering through agents yet?

The company remarks cited here point to convenience and trust as possible barriers. Xu said ordering food is typically done through DoorDash’s app, while Amazon’s Andy Jassy said outside-agent shopping was not yet more convenient than buying directly. The available information does not identify a single cause for the low use.

What should readers watch for next?

Look for companies to report agent transactions with a clear time period, definitions and completed-purchase counts. DoorDash has not announced a timetable or a current revenue-sharing arrangement for agent orders.

Source: ThorstenMeyerAI.com

You May Also Like

The labor share. Is value really moving from labor to capital? The data isn’t on anyone’s side yet.

Recent data shows stable aggregate labor share over 70 years, but early signals suggest marginal shifts. The debate on value transfer remains unresolved.

QAtrial: Compliance That Shows Its Work

QAtrial introduces a provenance-first AI tool for regulated life sciences, supporting compliance with 21 CFR Part 11 and EU Annex 11.

Trade voice copilo

A new voice copilot tool is being tested for small trades businesses to streamline job notes and invoicing, aiming to reduce admin hours and improve cash flow.

The Impact Of OpenAI’s AI Expansion In Brazil’s Tech Scene

OpenAI established a local team in São Paulo on August 27, expanding its presence and collaborations in Brazil’s growing AI market.