The “agent gap”: big companies are pulling ahead

McKinsey's latest State of AI survey shows enterprises scaling AI agents while smaller firms stall. The difference isn't budget — it's ownership.

McKinsey's latest State of AI research has a number worth sitting with: the share of large enterprises scaling AI agents in at least one function jumped from 27% to 40% this year. For smaller organisations, the figure barely moved — stuck around 22%.

It's not because agents need enterprise budgets. The same survey found roughly a third of organisations have skipped buying at least one piece of software because they could build the capability internally with agentic tools. The technology is more accessible than it has ever been.

The difference is ownership. Big companies have someone whose actual job is making AI stick — choosing the right first project, wiring it into real workflows, training people, and staying with it past the demo stage. Most midsize businesses don't, which is also why so many AI projects fail: industry analyses this year keep landing on failure rates around 80%, mostly for organisational reasons, not technical ones.

The fix isn't hiring a head of AI. It's borrowing one — someone who picks the boring-but-valuable first automation, proves it in weeks, and hands your team something they own.

That's the shape of every engagement we run. If you'd rather be in the 40% than the 22%, that's a conversation worth having.