Uber (UBER) AI Cart Builder: Carts Often Twice the Size of Non-AI Carts

On its 2026-08-05 earnings call, Uber Technologies, Inc. (UBER) said carts built with its AI Cart Builder are often twice the size of non-AI carts, and that about three-quarters of rides happen through a personalized destination suggestion. The company did not connect either reading to revenue or gross bookings [1].

What Uber's AI personalization does

Uber runs three businesses: Mobility, Delivery and Freight. Mobility matches riders with drivers. Delivery covers restaurant delivery and grocery delivery through Uber Eats. The platform recognizes revenue as a take rate on the gross bookings of each order, and it also earns advertising revenue from merchants.

The AI application here is a personalized service discovery and ordering capability. Models read a user's order history, real-time signals and behavior across Uber's businesses to predict where the user wants to go and what they want to buy. For rides, it pre-fills the destination from ride history, so the user does not have to type it. For food, it decides restaurant ranking, which deals are shown and how ads are placed. For groceries, Cart Builder lets a user take a photo of a dish or write down a recipe; the AI then talks with the user and builds a cart that can be ordered directly. The capability sits in the main path from opening the app to placing an order, and it serves Uber's two core businesses, Mobility and Delivery.

How the capability evolved, 2024 to 2026

The trail starts with the 2024-02-07 earnings call. CEO Dara Khosrowshahi said teams were already using AI algorithms to run promotions, targeting the right offer at the right price to the right user in each situation. He said only that this directionally raises order frequency and gave no performance figure [2].

On 2025-05-07, personalization moved from promotions into search. Uber began using larger models in restaurant and grocery search to adjust restaurant ranking and how deals are displayed. Management said at the same time that applying these models to the consumer experience was still very early [3].

On 2026-05-06, the company gave its first quantified usage reading: AI algorithms had predicted and preselected the destination for about three-quarters of rides. The same call mentioned building a shopping cart from a photo, but only as a feature description [4].

On 2026-08-05, photo-to-cart returned under the name Cart Builder and carried an effect reading for the first time. Management also repeated the destination-suggestion coverage figure [1].

Reading the two latest figures

The two latest readings measure different things and should be read separately.

On Cart Builder, Khosrowshahi said consumers who use it love it, and that carts built with it are "often" twice the size of non-AI carts [1]. That is a multiple of 2.0x, but the company did not say whether "cart size" means item count or dollar value. It also did not disclose how many people use the feature, the sample, or the measurement period. The comparison covers only users who chose to use the feature, and they may already be people who place large orders. The comparison has no control group.

On destination suggestions, management said the system guesses where a rider is going from ride history, and when it guesses right the rider does not need to type anything [1]. The "about three-quarters" figure shows how widely the feature reaches into the main ride-booking flow. It does not mean the feature generated additional trips. The wording differs from the 2026-05-06 statement, and neither statement gives a region or period, so the two figures cannot be used to calculate a change.

Possible path to Delivery gross bookings

The most likely financial channel is Delivery gross bookings. If Cart Builder makes more carts larger, the average order value of restaurant and grocery orders would rise. With order count unchanged, that would lift Delivery gross bookings, which convert into revenue through the platform's take rate. Destination pre-fill mainly removes typing steps when booking a ride, so its effect on Mobility gross bookings is more indirect.

Uber has not linked the cart reading or destination prediction to any change in gross bookings, revenue or profit, and it has not disclosed what share of orders Cart Builder covers. Delivery gross bookings also move with order volume, pricing, promotions, merchant supply and currency. For now this chain is only a possible, directionally related path.

What is confirmed and what is not

What can be confirmed is that this personalization capability has entered the main flows for booking rides and shopping for groceries. Destination suggestions cover most rides, and Cart Builder produces larger carts among the people who use it. Its contribution to average order value and gross bookings has not been quantified separately. Only if Uber discloses the share of orders that go through Cart Builder and its contribution to Delivery average order value could investors judge whether the capability has turned into incremental revenue.

Application assessment

• Personalized Service Discovery and Booking | Business position: core business | Stage: limited production | Coverage: multiple businesses or regions | Value type: revenue growth

Sources

[1] Drillr · Uber Technologies, Inc. (UBER) · 2026-08-05 · Earnings call

"And the effect there is consumers love it, the ones who use it, but also the size of those carts is often twice the size of kind of non-AI built carts."

[2] Drillr · Uber Technologies, Inc. (UBER) · 2024-02-07 · Earnings call

"All of these occasions, there are different occasions that we can target the right person with the right offer at the right price. All of that now is algorithmically driven."

[3] Drillr · Uber Technologies, Inc. (UBER) · 2025-05-07 · Earnings call

"It starts in smaller ways, so for example we’re using larger models in terms of our restaurant and grocery search so that we understand more about the context of the consumer, we get to know the consumer more, and we’re able to surface better results, higher quality results in terms of search, in terms of the sort order of restaurants that we’re offering you or the promotions that we’re offering you as well."

[4] Drillr · Uber Technologies, Inc. (UBER) · 2026-05-06 · Earnings call

"three-quarters of the rides on Uber, you know, we have successfully actually predicted with AI algorithms where we think you're likely to go."