Businesses Rush to Hire AI Agents, But Most Aren't Ready for the Mess

Over the last year, AI agents — software that can take actions on its own, like answering emails or processing orders — have gone from a futuristic promise to an everyday workplace experiment. New research from Deloitte, KPMG, and Accenture all tells the same story: companies are adopting these agents fast, but they're struggling with the harder parts. Only 15% of organizations have rolled out AI agents at scale, and just 16% say their current processes are ready.

That gap matters because the expectations are huge. Deloitte found that 74% of leaders expect half of business processes to be redesigned around AI agents by 2030. That would reshape how millions of people work. Yet KPMG's survey of over 2,100 executives shows that even as adoption climbs, measurable return on investment remains limited. The companies that do see real value share one thing: they have clear accountability, strong oversight, and a handle on what AI actually costs.

There is good news buried in the numbers. Salesforce research shows that AI agents have tripled in number at the average organization — from 5 to 13 — and employees are using them three times more as trust grows. Agents also became smarter, with capability improvements of 350%, and they can now be created in under two days. Middle managers, who are often the people making AI work in practice, are mostly optimistic and feel personally responsible for how their teams use these tools.

The uncomfortable truth? The biggest barrier isn't technology. It's governance, training, and the human side of change. Companies that focus on who's accountable when an AI agent makes a mistake, and how to redesign jobs around them, are the ones pulling ahead. Everyone else is just adding more agents and hoping for the best.