The language used to describe just about any modern company is utterly predictable, with the same words (agile, resilient, empowered) used over and over again. Spending any amount of time with any of these firms however usually reveals that the reality is rather different.
Endless Slack channels overflow with questions that people half answer at best. Decisions that need to be made quickly get postponed for weeks on end. Teams end up building and often implementing the same feature because nobody knew what anyone else was doing. Meetings multiply as team leaders panic about whether or not everyone understood what happened at the last meeting.
It’s easy to blame management consultants for the mess, so let’s do just that. They have spent the past decade preaching the virtues of flat organisations that strip out the layers in pursuit of quicker decisions, lower costs, and teams closer to customers. What they forgot to mention is that this flattening also strips out coordination capacity: the gritty, unglamorous work of figuring out what matters, when it matters, and who should do it.
Plenty of firms now realise, however, that they have been fooled, and are rebuilding the connections they have spent years removing. The middle manager, in other words, is back.
The case for going flat
The original, agility-first argument made perfect sense. Extra layers bred office politics. Hierarchies ballooned costs and kept talented people far from the customers who mattered. Get rid of it all and you’d unleash speed and creativity.
It worked brilliantly under specific conditions. Start-ups with a single product, a clear mission, everyone in the same room. High-clarity environments where informal networks could substitute for formal structure. But there was a mistaken assumption buried in all this, that alignment would just happen, and that shared context, physical proximity and mutual trust would somehow materialise without anyone working at it.
The Covid-19 pandemic and the explosion of remote work revealed the limitations of the flat model. Offices mask dysfunction through nothing more complex than human interaction. A short chat with someone while waiting for the kettle to boil solved any number of issues. Work from home removed that interaction and introduced the need for complex systems that immediately failed.
Research on technology workers during the pandemic found productivity fell by eight to 19 per cent despite people working longer hours. Time spent coordinating surged whilst uninterrupted work time collapsed.
Then there’s artificial intelligence, which is speeding everything up. McKinsey reckons AI agents handled over 60 per cent of repetitive workflows last year. This is all well and good until you realise that faster execution has the potential to multiply collision points. Teams can now ship in a day what used to take a week, or a month. But if those teams are all doing the same thing, without synchronisation, you end up with more rework, more misalignment, and a huge amount of wasted effort. Speed without coordination is just expensive chaos.
Meanwhile, the organisation’s interfaces keep multiplying. More cross-functional work. More compliance requirements. More vendor ecosystems. More data dependencies. Teams are autonomous only until their work touches someone else’s work. Which happens approximately always.
What the new middle manager actually does
The middle manager creeping back into corporate life bears little resemblance to the supervisor-by-presence model everyone rejected. The old version monitored attendance and relayed messages up and down the management line. The new version does things that can’t be automated or delegated away.
Sensemaking comes first. Someone has to translate strategy into ‘what we actually do this week’ and maintain shared context across distributed teams. Then prioritisation, which mostly means deciding what not to do and reconciling competing goals. Do we optimise for speed or for risk? Growth or reliability? Somebody has to take that decision.
Interface management is the third function. Owning the hand-offs between product, engineering, risk and operations. Reducing the invisible friction between teams. Decision design is fourth: setting who decides what, when, using which inputs. Building escalation paths that don’t turn into escalators to nowhere.
Finally, capability building. Coaching people, developing their judgment, spotting skills gaps before they become crises, redeploying talent before attrition does it for you.
The new middle manager is less foreman and more air-traffic controller.
The telltale symptoms
Companies that removed middle management without replacing the function tend to develop predictable patterns. Alignment meetings breed like rabbits. Teams duplicate work because nobody shares what they’ve learnt. Decisions happen everywhere, which means they happen nowhere. Strategy lives in slide decks whilst delivery becomes local optimisation. Accountability blurs because outcomes are shared, so responsibility becomes optional.
The tricky bit is that the output metrics often look fine. Activity stays high. Outcome metrics lag by months. By the time problems show up in customer experience or reliability or time-to-value, the damage has become cultural.
A survey of 15,000 workers across ten countries last year found that 44 per cent of Americans said their company had sliced away management layers in 2024. Yet hiring data from 2025 shows stronger demand for managers, senior managers and directors than for individual contributors.
Intelligent layering versus more bureaucracy
The solution isn’t just as simple, however, as rehiring managers, it’s adding back clarity. Intelligent layering means fewer levels but sharper decisions, local autonomy paired with global coherence. Teams own execution whilst managers own the interfaces between teams. Standards where they matter, flexibility everywhere else. Accountability by design, not by accident.
A workable model runs on three layers. Teams build, ship and learn. Orchestrators in the middle coordinate, prioritise and resolve trade-offs. Executives at the top set direction, allocate capital and shape culture. Each layer sticks to what it does best.
The AI twist
Artificial intelligence won’t kill off middle management, but it will make hiring the right middle managers, and using them correctly, more important. AI now brilliantly handles administrative tasks such as reporting, status updates, meeting summaries, and documentation. That frees up time for the judgment-based work such as making trade-offs, coaching people, and improving the quality of decisions taken. Better analytics also expose weak coordination faster, which sounds helpful until you realise it just makes the pain more visible.
The danger is using AI to accelerate noise instead of reducing coordination costs. If managers use it to generate more dashboards and more status pings, the organisation drowns faster.
IBM reported saving 3.9 million hours in 2024 by automating repetitive tasks. It also, however, crucially points out that human glue remains essential for holding workflows together. AI handles the mechanics. People handle the meaning.
What firms should actually do
Gartner predicts that three-quarters of organisations will this year face measurable productivity losses if they don’t address hybrid work complexity. The next productivity leap won’t come from cutting managers. It will come from improving decision quality and coordination throughput.
For executives, that means treating middle management as a capability engine, not overhead. Invest in selection, training, career paths, proper incentives. For operations leaders, it means mapping value streams and interfaces to work out where the middle layer actually earns its keep. For middle managers themselves, the brief finally makes sense. Your edge is judgment, not supervision.
The future organisation isn’t flat. It’s layered like decent infrastructure: mostly invisible, but unquestionably essential. To many companies tried removing the middle. It turned out it was holding the building up.
Photo: Dreamstime.
