Companies wait for AI productivity gains to scale

Companies are spending heavily on AI in hopes of lowering costs and raising output, but the payoff remains uneven. Software engineer Iren Azra Zou says Anthropic’s Claude Code lets her finish some tasks in a day that once took as long as a week, while Amazon data scientist Sarthak Gupta says AI is adding work as he builds pipelines, integrates tools, and shifts workflows into new systems.

Executives are increasingly linking AI with productivity. AlphaSense found the terms appeared together in 637 earnings calls in the second quarter, up about 25% from the same period a year ago. Yet a National Bureau of Economic Research working paper found roughly 90% of firms actively using AI reported no productivity impact over the prior three years, based on a survey of nearly 6,000 executives.

Economists and business leaders say the main obstacle is scaling isolated gains into companywide improvements. McKinsey’s Alexander Sukharevsky described a persistent gen AI paradox, while Uber COO Andrew Macdonald said greater AI use has not directly translated into more useful consumer features. Mark Zandi of Moody’s expects any economic lift to arrive slowly, possibly not until the late 2020s, or early 2030s.

AI may follow the path of spreadsheets and email, becoming a foundational workplace tool only after companies redesign processes around it. For now, businesses are still in the middle stage, trying to prove that individual efficiency can become durable growth.