|
Many corporate leaders say they’re getting value from AI as their spending on it skyrockets, but hype is outpacing reality in plenty of cases. On a podcast over the weekend, for instance, Uber Chief Operating Officer Andrew Macdonald said the ride-hailing company isn’t seeing a clear increase in productivity from using AI coding services despite their use by its engineering teams. That has prompted executives to discuss how to get a handle on token consumption costs, he said. “If you‘re not actually able to draw a direct line to how much useful features and functionality you’re shipping to your users, [the costs become] harder to justify,” Macdonald said on the podcast. It’s the latest sign that Uber is having a rocky transition to AI-powered coding. Last month, Chief Technology Officer Praveen Neppalli Naga told us his company’s surging usage of products like Claude Code had caused it to blow through its entire AI budget just a few months into the year. Uber isn’t alone. Many other firms are struggling to navigate Anthropic’s shift to charging customers based on token consumption, which has made it harder for them to gauge costs in advance. For now, customers are eating the higher costs, ROI be damned. Other AI projects outside coding have simply fallen flat or created unintended consequences. Take Pizza Hut, which in 2023 and 2024 launched an AI-powered ordering system, Dragontail, that was supposed to speed up deliveries for its restaurants in the northeastern U.S. Instead, Dragontail led to much longer delivery times and thus more frustrated customers, according to a lawsuit filed by a Pizza Hut franchisee that runs more than 100 of the company’s restaurants in the northeast U.S. The franchisee is seeking $100 million in damages on the grounds that Pizza Hut forced it to use the system. After Pizza Hut rolled out Dragontail, the system allowed DoorDash’s delivery staff to see operational details they weren’t previously privy to, like the status of pizzas as they were being cooked and whether a customer had left a tip, according to the filing. Armed with this knowledge, DoorDash delivery staff would wait for multiple orders to be ready instead of delivering pizzas as they came out of the oven, as they had previously done. In other cases, delivery staff would decline to deliver orders to customers who hadn’t tipped. The result was that some customers received cold pizzas and left negative reviews, according to the filing. Meanwhile, Starbucks, which nine months ago launched an AI-powered inventory tool that employees used to automate the counting of milk and beverages in stores, has scrapped the tool after finding it would often miscount and inaccurately identify items, Reuters reported. Starbucks reportedly spent years developing the tool and saw it as a way to help the company get a better handle on its supply chain operations and deal with frequent product outages. Starbucks, in a statement via email, said it has “moved to a single, consistent process” for counting inventory and that the change “reflects being disciplined about where automation adds value.” “We test ideas in our coffeehouses, listen closely to partner feedback, and make changes to deliver a better, more consistent experience,” a Starbucks spokesperson said in a statement via email. Some AI providers have struggled to turn their marketing of AI products into reality. Salesforce, for instance, last year featured Williams Sonoma using Agentforce to run a customer support phone line, and Finnair (a Finnish airline) using the product to help customers rebook flights, but the companies told Bloomberg last week that they aren’t able to use the software that way. The Information previously reported in detail that Salesforce has backtracked from earlier claims about the relative ease of setting up AI inside enterprises, and the company has reported better financial results with Agentforce since then.
|
Gift Subscription
