Generative AI stalls as businesses struggle to see real gains

Summary

A sweeping Reuters investigation finds that, three years after ChatGPT ignited the generative AI boom, most companies still aren’t seeing meaningful profit or productivity gains from their deployments. Surveys by Forrester and BCG show only 5–15% of executives reporting improved margins from AI, with many now delaying roughly a quarter of planned AI spending into 2026 as projects prove harder to operationalize than promised. Executives describe a "jagged frontier" where large language models excel at complex math or coding but routinely fail at simpler business tasks, like summarizing a 100‑page safety manual or giving blunt product recommendations, forcing firms like CellarTracker and Cando Rail to scrap or heavily rework pilots. OpenAI, Anthropic and Google are responding by embedding applied AI teams more deeply with customers, betting that hands‑on co‑development and domain‑specific agents—rather than generic chatbots—will finally turn massive infrastructure bets into real returns. For now, though, the story is one of expectations reset: AI remains a priority, but boards are demanding clearer roadmaps, better data plumbing, and more human oversight before signing off on yet another nine‑figure AI budget. ([reuters.com](https://www.reuters.com/business/business-leaders-agree-ai-is-future-they-just-wish-it-worked-right-now-2025-12-16/))