Only 11% of Companies Have Reached AI Reinvention
McKinsey surveyed 750 leaders and employees. Organisational readiness predicted value capture about twice as strongly as individual readiness, which says a lot about where most AI programmes are stuck.
McKinsey surveyed 750 employees and leaders and found that 11% of leaders say their organisation has reached what the firm calls the reinvention stage. Nearly 90% are still doing one of the two earlier things: helping employees use AI, or automating workflows that already exist.
The report's frame is three stages — enablement, automation, reinvention — and its central finding is that the value is concentrated in the third one.
The gap that actually predicts value
70% of respondents said they personally felt ready to adopt and use AI. Only 27% of leaders thought their organisation was ready to make the changes an agentic future demands.
McKinsey then did the useful thing and asked which readiness mattered. Organisational readiness accounted for 48% of the difference between leaders who reported capturing enterprise value and those who didn't. Personal readiness accounted for 25%.
That ratio is the finding. Individual capability is roughly half as predictive as whether the organisation around that individual can absorb what they produce.
Redesign beats enablement
Three numbers from the report, all comparisons against organisations that didn't do the thing:
- Leaders were 5.3x more likely to report enterprise value capture when workflows were redesigned rather than left unchanged.
- Leaders with highly AI-fluent teams were 3.9x more likely.
- Leaders who received support and training to build new skills were 3.3x more likely.
The mechanism McKinsey describes is straightforward. Making employees faster does not improve business performance unless the freed-up capacity gets redirected to something that matters. Absent that redirection, you have bought time back and then spent it on the same work.
Reading this from an enterprise seat
Most AI programmes I've seen are structured as enablement programmes — licences, training, a centre of excellence, an internal champions network. All of that is stage one, and this data says stage one on its own doesn't pay.
The uncomfortable implication is that the constraint isn't tooling or talent. It's whether anyone has the authority to redraw a process across function boundaries, and whether the operating model can survive it. That's not a technology decision, and it's not usually the CIO's to make alone.
If your AI programme has no workstream that changes who does what, it is an enablement programme wearing a transformation label.