McKinsey and HBR research consistently identifies middle managers as the primary bottleneck in AI adoption. The standard organisational response is to route around them: provide AI tools and training to frontline practitioners, report outcomes to senior leadership, and give middle managers updates rather than participation in design. This produces 1.5x worse outcomes than including middle managers explicitly in the workflow design process. The label of bottleneck mischaracterises the role. Middle managers are not obstructing AI. They are defending the governance functions they currently fulfil, functions that the AI deployment, as designed, has removed without replacement.
Middle managers resist AI that bypasses them because AI, in most deployment designs, removes the judgment exercise that is the primary source of their authority and value. When AI produces a recommendation and the workflow specifies that the practitioner follows it, the middle manager's role in reviewing and refining practitioner judgment disappears. The governance-aware response is to design middle managers in as explicit decision-rights holders: the named human checkpoint, the escalation owner, the person who reviews AI-informed outputs before they affect customers. This converts them from potential resistance into governance infrastructure. The manager who has a defined governance role in the AI workflow has a stronger accountability position than they had before the deployment. The manager who has been bypassed has a weaker one and will act accordingly.
For each AI workflow being designed, identify the middle manager whose function it sits within and assign them a specific governance role: a named checkpoint, a defined escalation point or an explicit quality review responsibility. Give them the criteria that govern that role. This is not a change management concession. It is a governance design requirement. The manager who knows exactly what their AI governance function is will champion the workflow. The manager whose judgment has been replaced without a replacement role will find reasons for it to fail, and they will be correct to do so, because the governance gap their removal created is real.
setmode.io is a 14-module programme that closes the gap between AI adoption and organisational transformation. Every engagement produces named deliverables that form an AI-enabled operating model.
Start a conversation