The Company That Replaced Its Managers With AI

‍

When Block announced it was cutting more than 4,000 employees — roughly 40% of its workforce — the story could have been told as another tech industry contraction. CEO Jack Dorsey framed it differently. In a public post co-authored with other executives, he argued that AI can now perform the core function of middle management: routing information up and down an organizational chart. That reframing changes the nature of the story entirely.

‍

Block is not simply running leaner. It is making a structural argument about how companies should be built in an era when AI can process, prioritize, and direct the written record of organizational life. The company is reorganizing around three roles — builders, problem-owners, and player-coaches — and treating coordination itself as a software problem. Whether that bet holds is a question every operations leader should be watching carefully.

‍

What Block Actually Did — and Why It Matters

‍

To understand Block's move, it helps to understand what made it structurally possible. Block is a remote-first company. Nearly every decision, escalation, and status update flows through written channels — Slack messages, project management tools, recorded meetings. That digital paper trail is not just a byproduct of remote work. It is the operational substrate that AI coordination systems need to function.

‍

In a traditional office, a middle manager absorbs information through hallway conversations, body language, and informal context that never gets written down. That information is real, but it is invisible to software. Block's model works in part because its information environment was already legible to machines before the restructuring began.

‍

This is not a typical cost-cutting story dressed up in AI language. It is a public thesis — one Dorsey is willing to test at scale — about whether digitally native organizations can eliminate an entire structural layer by treating coordination as infrastructure rather than headcount. The practical implications extend well beyond one fintech company.

‍

The Case for AI as an Organizational Layer

‍

Middle management's core function in large organizations has always been information translation. Executives set direction; teams execute. Managers convert strategy into tasks, surface ground-level signals upward, and filter the noise that would otherwise overwhelm both ends of the hierarchy. That translation process is valuable — but it is also slow, lossy, and prone to distortion introduced by each human intermediary.

‍

AI systems can perform a version of this function with less latency and more consistency. They can parse structured communications, identify blocked work, flag misaligned priorities, and route decisions to the right person without the delays introduced by a manager who is in three other meetings. In that narrow sense, Dorsey's argument has genuine merit.

‍

The logic echoes earlier arguments about flat organizations — companies like W.L. Gore and early-stage startups that operated without formal management layers. What is new is the enabling infrastructure. Previous flat-organization experiments relied on strong cultural alignment and small team sizes to hold coordination together. AI offers a potential alternative: not culture as the substitute for hierarchy, but software. Whether software is a sufficient substitute is the question Block is now answering in real time.

‍

Where the Model Breaks Down

‍

The honest version of this argument acknowledges what AI coordination systems cannot do. Management is not only information routing. It includes performance feedback delivered with care, conflict resolution between colleagues who have stopped communicating, psychological safety built over months of consistent behavior, and career development that requires a human to advocate for someone else's growth. These functions do not live in the written record. They live in the relationship.

‍

There is also a documented problem with how AI systems handle disagreement. Researchers have noted a pattern called AI sycophancy — a tendency for large language models to affirm user input rather than challenge it. A manager who never pushes back is not a good manager. An AI system structurally inclined toward agreement is a poor substitute for someone who will tell a team lead their plan has a flaw.

‍

The history of radical flatness without AI offers a cautionary note as well. Zappos' experiment with holacracy produced well-documented coordination failures and significant voluntary attrition. Valve's flat structure, celebrated in its employee handbook, has faced persistent criticism for creating informal power hierarchies that went unacknowledged and unmanaged. Removing formal structure does not eliminate the need for structure. It just makes the structure harder to see and harder to fix.

‍

It is also worth noting that Block's 4,000 departures were not limited to middle managers. A reduction of that scale touches institutional knowledge, cross-functional relationships, and organizational memory in ways that are difficult to model in advance.

‍

What Business Leaders Should Do Right Now

‍

Block is functioning as a live case study. The broader business world does not need to replicate its model to learn from it. The more useful exercise is asking a precise question about your own organization: how much of your management layer is doing information routing, and how much is doing something that genuinely requires human judgment?

‍

The ratio varies widely. A software company with fully distributed teams and documented workflows will look very different from a manufacturer with unionized labor and in-person operations. AI coordination tools are not universally applicable. The companies most likely to benefit share Block's profile: remote-first, heavily digitized, with workforces comfortable operating asynchronously.

‍

For those who see partial overlap, a few practical steps hold regardless of where you land on the larger debate. Audit your existing management layer with honesty about what it actually produces. Invest in the digital infrastructure — written decision logs, structured async communication, documented workflows — that makes AI coordination viable before you try to deploy it. And resist the temptation to conflate workforce reduction with AI transformation. Cutting headcount ahead of effective AI deployment does not create efficiency. It creates operational risk with a technology-forward label attached.

‍

Watch Block's performance over the next 12 to 18 months. If revenue holds, product velocity increases, and employee retention stays stable, the model earns credibility and others will follow with more confidence. If coordination breaks down or talent exits accelerate, it will tell the business world something equally important: that AI can route information, but it cannot yet replace the judgment, relationships, and accountability that make management worth having in the first place.