The emotional labour of AI adoption is landing on middle managers
Debbie Green explores the pressures AI-driven change places on middle managers and the support they need to navigate it.

There’s a particular kind of tiredness that comes from having to look calm for everybody else, and many middle managers know it well. They hold deadlines, customer expectations and performance concerns; they translate leadership decisions, absorb employees’ questions and try to keep teams motivated, even while they’re still processing the change themselves. Then AI-driven transformation gets added to the mix.
The conversation about agentic artificial intelligence (AI) is often framed around productivity: what can be faster, automated, scaled. Far less attention goes to the human energy required to redesign work while continuing to deliver it. McKinsey’s recent research on ‘How to cluse the agentic adoption gap’ identifies middle managers as the group most anxious about AI-related workplace change.
The report describes four fears that commonly accompany AI adoption: being exposed as someone who doesn’t know, being accountable for technology one can’t fully control, moving towards an unclear destination, and losing the expertise professional identity was built on. These aren’t abstract concerns; they show up in the body, mood and behaviour of managers every day.
The emotional labour in the middle
Middle managers often become the emotional shock absorbers of change. An employee shares a fear about redundancy, a senior leader wants visible momentum, a new AI system gives an unreliable answer. The manager absorbs each pressure, translates it and tries not to pass the anxiety onwards.
This emotional labour is real work, but it’s rarely recognised in workload planning. Over time, the demand to contain everyone else’s uncertainty can disconnect managers from their own needs. They postpone breaks, work longer to master unfamiliar tools and hide anxiety to protect their credibility. Wellbeing support must start by acknowledging this reality. A meditation app cannot compensate for an impossible role.
Learning adds to this if organisations don’t make room for it. Managers expected to complete courses, experiment with tools and coach their teams, all alongside existing workload. They receive a clear message: adapt, but do it in your own time. That produces a familiar cycle: too busy to learn properly, so confidence stays low; low confidence invites more pressure; more pressure reduces psychological safety further. To interrupt it, learning must be treated as protected work, not something squeezed around the edges.
Anxiety is information










