Middle management has had a complicated public reputation for decades — the Dilbert-cartoon caricature of bureaucratic obstruction, unnecessary hierarchy, and the pointless coordination layer that sits between people doing real work and the executives making decisions. This reputation is unfair in its blanket application but contains enough truth to have shaped the way technology companies in particular have approached organisational design, with a consistent preference for flat structures and direct reporting relationships over hierarchical layers.
The predictions have been dramatic. But is middle management actually disappearing — or just changing? AI agents can perform the coordination, information routing, and process management functions that middle managers currently provide, eliminating the need for the human layer. The argument is interesting, and partially right, and importantly wrong in ways that matter for how businesses should actually think about the AI impact on their organisational structures.
What Middle Managers Actually Do
The Dilbert caricature focuses on the coordination bureaucracy aspects of middle management — the meetings, the status updates, the permission-granting and approval-seeking. These functions are real, often over-developed, and genuinely reducible by AI-mediated coordination and workflow automation. If a significant part of a manager’s role is routing information, scheduling, and managing task tracking, that part is automatable.
But the research on middle management — which is more positive than the cultural caricature — identifies functions that are considerably harder to automate. Middle managers serve as translators between senior strategy and frontline execution, adapting abstract direction to specific operational contexts in ways that require deep understanding of both. They identify the early signals of problems that aren’t yet visible in dashboards: the team that is burning out, the process that is producing the right metrics but the wrong outcomes, the customer relationship that is more fragile than the retention numbers suggest. They develop people — coaching, mentoring, creating growth opportunities, building the next layer of leadership capability. And they absorb and manage ambiguity so that frontline employees can work with greater clarity than the senior leadership’s own uncertainty would permit.
What AI Can and Can’t Replace
Networks of AI agents — systems where multiple AI tools hand off tasks to one another automatically — are genuinely capable of automating significant portions of process management, information routing, and routine coordination. The workflow orchestration, project status synthesis, and task prioritisation functions that consume significant management time in administrative-heavy organisations are all legitimate targets for AI automation. Companies that have deployed AI workflow tools have seen meaningful reductions in coordination overhead and improvement in the speed of information flow.
The people functions — coaching, development, sensing of human dynamics, building trust, managing the interpersonal complexity of teams under pressure — are not being replaced by AI systems in any near-term realistic sense. These functions require the kind of contextual judgment, emotional attunement, and relationship depth that current AI systems don’t have. More importantly, people being managed often need to feel understood and valued by someone who has genuine stake in their situation — a need that AI-mediated management, however technically capable, is unlikely to satisfy.
The Realistic Restructuring
The honest prediction isn’t the death of middle management but a significant change in what middle managers do and how many you need for a given level of output. AI automation of coordination and process management will reduce the portion of management time spent on those functions, which will either reduce headcount at the management layer or free management time for the higher-value people and strategic development work that AI doesn’t perform.
The management roles that are most at risk are those whose primary value is information routing and process oversight with minimal human judgment or development content. The management roles that are most durable are those where the primary value is context-sensitive human judgment, relationship depth, and the development of people. That’s not a universal story of displacement — it’s a restructuring of function that will look different organisation by organisation, sector by sector.
