The evidence is stark. During the week of Halloween, AI agents fulfilled only four Uber rides, along with about 500 Uber Eats orders, according to Uber executives. That compares with roughly 320 million trips and orders handled across Uber and Uber Eats during the same period, TechCrunch reported.
The numbers offer one of the clearest real-world tests yet of the commercial claims surrounding AI agents. Technology companies are building assistants that can move beyond answering questions and take actions such as shopping, booking travel and ordering food. But Uber’s results indicate that consumer adoption remains far behind the industry’s expectations.
Read Amazon Blocks Meta’s Muse, Turning AI Shopping Into a Platform Fight
On Uber’s third-quarter earnings call, Chief Financial Officer Balaji Krishnamurthy said agents had fulfilled four rides and approximately 500 Uber Eats orders during the comparison period, according to the earnings-call transcript. Chief Executive Dara Khosrowshahi described the traction as “very, very early and very limited,” while saying the company expects the number of major AI assistants in the market to rise above 10.
That response is strategically important. Uber is not rejecting agents. Its prepared remarks for the third-quarter results said AI could help users discover services and complete transactions, while also emphasizing that Uber’s own app is already fast. The company said booking a ride takes about 50 seconds and reordering a favorite meal takes roughly 90 seconds.
That creates a difficult business case for agents. If a customer can open an established app, select a known destination or meal, confirm payment and complete the transaction in less than two minutes, an intermediary assistant must offer more than novelty. It must save meaningful time, improve the outcome or solve a problem the app cannot handle well.
Convenience is not automatically a moat
The critical argument is that the agentic web may be adding a layer of friction to transactions that platforms have already optimized. Users may need to grant permissions, authenticate accounts, confirm payment information and review an order generated by a system that does not have the same history or trust as the native app.
TechCrunch’s report highlighted that contrast between Uber’s massive transaction volume and the tiny number attributed to agents. It also presented Uber’s fast native experience as a possible explanation for the weak usage.
That criticism matters because the strongest consumer technology businesses often win by removing steps. An agent that introduces another interface, approval process or point of failure may not be a superior distribution channel. For Uber, the challenge is especially pronounced because its app already knows the customer’s locations, payment preferences, previous orders and frequently used services.
The same issue applies to platform economics. If an agent sits between a customer and Uber, the company may need to manage new authentication systems, fraud risks, commissions, customer support responsibilities and disputes over who owns the customer relationship. More demand would make those costs worthwhile. Four rides and 500 food orders do not yet demonstrate that the volume will arrive.
Where agents may gain ground
The weak numbers do not prove that agents have no commercial future. They show that routine, familiar transactions may be among the least attractive starting points.
An agent could be more useful when a customer has an unfamiliar destination, needs to compare several transportation options, is coordinating travel for a group or wants to combine multiple services. Food ordering may offer more room for assistance when a user has complicated dietary restrictions, wants to compare restaurants or is trying to plan meals rather than repeat a previous order.
DoorDash Chief Executive Tony Xu made a similar case from a different position. In DoorDash’s third-quarter earnings call transcript, Xu said third-party AI agents had generated only several hundred orders in total, but argued that agents could produce incremental demand, particularly for longer-consideration purchases or orders initiated proactively.
That view frames agents less as replacements for existing app behavior and more as potential demand generators. The distinction is important. An assistant may not persuade a customer to delegate a familiar Uber booking, but it could prompt a purchase the customer had not planned to make, compare alternatives across services or complete a multi-step request that would otherwise be abandoned.
For Uber, the competitive question is therefore not simply whether agents can book rides. It is whether they can expand the market, increase order frequency or reach customers outside the company’s own application. Khosrowshahi said AI companies are already discussing integrations with Uber, suggesting that the company wants access to those channels without surrendering control of its core platform.
The early data gives Uber leverage in those negotiations. The company can participate selectively while its native app remains the fastest route for ordinary use. AI developers, meanwhile, must prove that agents can deliver incremental value rather than merely redirect transactions that customers already complete efficiently.
For now, the agentic web remains a strategic promise, not an established consumer habit. The next stage of competition will be decided by whether assistants can create new demand and handle genuinely complex tasks, rather than whether they can reproduce a 50-second booking flow.
Rebeca Smith is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Rebeca Smith's usual length and in Rebeca Smith's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.