McKinsey: 30% of agentic AI adopters saw productivity decline

A widely syndicated McKinsey finding should give business owners pause. Roughly 30% of companies deploying agentic AI reported lower productivity, not higher productivity. That matters because agentic tools are often presented as a quick route to automating work across teams. The evidence suggests the opposite can happen when the work, controls and ownership around the tool aren’t clear. You can read the reported McKinsey finding here.

The practical lesson is not to avoid agentic AI. It’s to treat it like an operating change, not a software purchase. Before putting an AI agent into a business process, be clear on the task it owns, what decisions it can and cannot make, who checks its output, and how you’ll know whether it is saving time or creating rework. If nobody owns those questions, productivity can fall while activity appears to rise.

Start with a narrow, repeatable workflow where the current process is understood. Measure the baseline first. Then test the agent against a defined outcome, rather than asking whether it feels useful. This is the kind of thing we build into an AI command centre, giving leaders a clearer view of where AI is helping, where it needs oversight, and where it may be adding friction.