In 2025, Shopify CEO Tobi Lütke sent a memo to all employees that became widely circulated in business media. The message was blunt: before any team could request new headcount, they were required to demonstrate that they had explored whether AI could do the job first. "Reflexive hiring," as he put it, was over. Headcount would only be approved for work that AI genuinely could not handle.
Klarna, the buy-now-pay-later company, reduced its global workforce by roughly 700 people and publicly credited AI as the replacement for much of that capacity. The story is instructive in two directions: their AI did handle the equivalent of 700 full-time agents' worth of inquiries — but CEO Sebastian Siemiatkowski later acknowledged the company "went too far," and Klarna began rehiring customer-facing staff after service quality declined. (references below) The Klarna case is less a blueprint and more a reminder that AI can absorb enormous transactional load, but the transition requires careful management of what gets lost in the process.
Duolingo restructured its contractor workforce in 2024, reducing content creation roles after deploying AI to generate the lesson content those contractors were producing.
These are not fringe cases. They are early signals of a structural shift that research institutions, consulting firms, and workforce analysts have been tracking with increasing alarm — and that every business owner needs to understand, regardless of company size.
What Middle Management Actually Does
Before asking whether AI can replace managers, it is worth being honest about what most managers actually spend their time on.
Across industries, management researchers consistently find that a substantial share of managerial hours goes to coordination and information-processing — not strategic judgment. McKinsey's analysis of generative AI's economic potential found that 60–70% of knowledge worker activities across occupations are automatable with current AI technology. (references below) That range applies acutely to middle management, where the core daily work looks like this:
- Aggregating information from multiple sources and producing summaries for senior leadership
- Routing requests, tasks, and information between teams
- Status reporting — tracking what is happening, who is behind, what is blocked
- Scheduling coordination across multiple stakeholders
- Drafting documents, emails, and presentations that summarize work done by others
- Running recurring meetings whose primary function is information sharing
None of these tasks require strategic judgment or relationship skills. They are information processing and coordination tasks — exactly the category where AI has achieved human-level or better-than-human performance in the last two years.
The management work that genuinely requires human judgment — mentoring individuals, navigating ambiguous decisions, building trust with clients, making strategic bets under real uncertainty — represents a much smaller fraction of most managers' actual hours. The rest is coordination overhead. And coordination overhead is precisely what AI systems are built to eliminate.
The Numbers: How Fast This Is Happening
The pace of this transition is faster than most organizations are prepared for.
Gartner's 2024 workforce research predicted that by 2026, 20% of organizations will have eliminated more than 50% of their middle management positions. (references below) That is not a projection for 2035. That is a prediction for next year — and given that Gartner is historically conservative in its timeframes, the actual pace may be faster.
Korn Ferry, one of the largest executive search and organizational consulting firms in the world, surveyed more than 15,000 professionals worldwide for their Workforce 2025 research. They found that 41% of companies have already reduced managerial layers — actively, not just through attrition. One in three of those companies said AI-enabled automation was the primary reason for the reduction. (references below)
The pattern that matters: Middle management exists primarily to handle coordination and information flow between individual contributors and senior leadership. AI handles both of those functions better, faster, and cheaper than a human layer. This does not mean managers add no value — it means the part of management that is irreplaceable (judgment, mentoring, trust) is a much smaller fraction of most managers' actual hours than organizations have historically needed to pay for.
The "Great Flattening": What Axios Documented
In 2024, Axios coined the term "the Great Flattening" to describe the organizational trend being reported across industries: companies removing layers between individual contributors and senior leadership, with AI handling the coordination work that those middle layers previously provided.
The pattern Axios documented was consistent across sectors — technology, retail, financial services, healthcare administration, and professional services. Companies were not just cutting headcount in layoffs. They were restructuring permanently — eliminating the management tier and replacing its function with AI systems that could do the information processing and routing at scale, without the headcount cost.
For large enterprises, this represents tens of millions of dollars in annual labor savings. For small and medium businesses, the dynamic is different but equally important: the question is not whether to eliminate management layers you may not have, but whether you can run a 30-person operation's output with 12 people by building the AI infrastructure that handles the coordination work.
What AI Can and Cannot Replace in Management
The "AI is replacing managers" framing needs nuance, because it is not universally true across all management functions. The honest breakdown:
AI is already doing this better than most managers:
- Real-time project status tracking and exception reporting
- Cross-team information routing and task assignment
- Performance data aggregation and pattern identification
- Meeting summaries, action item extraction, and follow-up tracking
- Demand forecasting and resource allocation recommendations
- Compliance monitoring and documentation verification
AI cannot reliably replace this (yet):
- Navigating interpersonal conflicts that require emotional intelligence and relationship context
- Making judgment calls in genuinely ambiguous situations with no clear precedent
- Building organizational trust and culture in a human way
- Mentoring individuals through career development in a meaningful sense
- Strategic decisions that depend on external context and tacit knowledge that has not been codified
The managers who are most at risk are those whose value is primarily in the first category. The managers who are most likely to remain — and whose roles will expand — are those who have developed genuine capability in the second category.
For business owners, this means the managers worth retaining and investing in are those who can do the judgment work. The coordination work should be moved to AI as fast as possible — not as a cost-cutting exercise, but because AI is genuinely better at it.
What This Means if You Run a Small or Medium Business
The Great Flattening is not primarily a story about Fortune 500 companies downsizing. It is a story about what a lean, AI-enabled operation can produce relative to a traditionally staffed one.
If your competitor has 40 people and you have 15 — but those 15 are running AI systems that handle the coordination, reporting, and information processing that their 40 people are doing manually — who has the structural advantage? The answer is obvious. And that scenario is not a 2030 scenario. It is available to build today.
The practical implication for SMB owners:
- Before your next hire, ask what AI could do instead. This is not about being cheap. It is about allocating human capacity to work that generates more value than the cost of the person.
- Redesign manager roles around judgment, not coordination. If someone on your team spends most of their week on status tracking, scheduling, and information routing — that is a sign those functions should be automated. Redirect their capacity toward decisions only they can make.
- Build coordination infrastructure before the scale problem hits. The worst time to realize you need AI coordination systems is when you have already hired a management layer to handle the coordination manually. Build it early, when the process is still simple enough to codify cleanly.
- Use AI-enabled management tools to stay flat as you grow. The goal is to grow revenue without growing headcount proportionally. That requires AI handling the operational overhead that would otherwise require more managers.
The Uncomfortable Truth for Business Owners
If you have managers — or if you are considering adding them — the honest question to ask is: what percentage of their time is on tasks that AI could handle today?
If the answer is more than half, you are paying for coordination overhead at a human salary. That is not a criticism of the person — it is a systems problem. The role was designed before AI could do those tasks. The role needs to be redesigned, not the person replaced.
The businesses that get this right will carry significantly lower operational costs into a period of intense competitive pressure. The ones that hire their way through the scaling problem — adding layers of coordination and management as they grow — will find themselves structurally disadvantaged against competitors who have built the same operational capacity on AI infrastructure at a fraction of the cost.
This is not a distant threat. Gartner's timeline is 2026. Korn Ferry's data shows it is already happening in 41% of companies surveyed. The businesses building AI operational systems now are the ones that will look inevitable in retrospect.
The question is not whether the Great Flattening is coming to your industry. It is whether you are building the AI layer that makes you the business that runs flat — or the one that is still staffing coordination overhead when your competition is not.