McKinsey: One in Four Middle Managers Report AI-Change Anxiety — Trust, Not Tech, Predicts Who Adapts

McKinsey’s “AI transformations run on trust” makes an unusually blunt claim for a strategy-firm publication: employee trust, not the technology itself, is the strongest predictor of whether an AI transformation actually captures value. The piece cites the 2026 Edelman Trust Barometer finding that 78% of employees globally say they trust their employer “to do what is right,” then layers on McKinsey’s own research showing roughly one in five employees across all levels report anxiety about AI-driven changes — rising to one in four among middle managers specifically. Employees with low trust in organizational support are 1.5 times more likely to report that anxiety than high-trust peers.

McKinsey’s framing — “technology may power AI transformation, but people determine whether it succeeds” — lays out four trust-building actions for leaders: offer a clear plan while still acknowledging uncertainty, get out into the organization for direct engagement rather than top-down messaging, invest in sustained capability-building and visible career pathways, and equip leaders at every level to build trust consistently, including handling workforce transitions with fairness and transparency.

The same week, BCG published its own CEO-facing piece on why AI transformations stall, naming “paralysis” as a distinct failure mode from simply moving too slowly — teams stop advancing not because leadership opposes the effort, but because they lack the clarity or alignment to act on nominal support. Read together, the two firms are converging on the same diagnosis from opposite ends of the org chart: BCG’s paralysis sits at the leadership layer, McKinsey’s anxiety sits with the managers actually absorbing the change — and both trace back to a trust and clarity gap rather than a capability gap.

For consulting engagements, the middle-manager number is the one worth sitting with: a quarter of the layer responsible for translating strategy into daily practice reporting real anxiety about the change they’re supposed to drive is a change-management problem, not a tooling one, and no amount of model capability closes it on its own.