Singularity companies: the end of hierarchical management in 12 months

Image: jurvetson — Openverse (by)

There’s something nobody says out loud in the boardroom.

Not because it’s wrong. Because it’s terrifying.

Hierarchical management — the kind taught in MBA programmes, optimised with OKRs, humanised with “team rituals” — is dying. Not in ten years. Not in five. In twelve to eighteen months, for every company whose core business is intellectual production.

What’s replacing it doesn’t have a settled name yet. I call it a singularity company: a structure where a single human orchestrates a network of autonomous AI agents to produce what used to require a team of 10, 30, sometimes 100 people.

This isn’t speculation. It’s a model I’ve been living every day for eighteen months. And what I observe convinces me of one thing: we’re not talking about an optimisation of work. We’re talking about the extinction of an organisational species.

---

Hierarchical management has 12–18 months left

Let’s be precise about what’s dying.

It’s not leadership. It’s not strategic vision. It’s not the human capacity to decide under uncertainty. What’s dying is the human infrastructure of coordination — middle management, sync meetings, reporting chains, approval workflows, weekly check-ins, RACI matrices.

That entire layer exists for one reason only: to transmit information between human brains that can’t talk to each other in real time, that forget, that misinterpret, that have egos, that take Fridays off.

Agentic AI doesn’t solve those problems. It eliminates them. An agent doesn’t need you to cascade a quarterly objective into monthly sub-goals broken down into weekly tasks. It receives an objective, accesses the context, executes, reports back. On a loop. Without a meeting.

The question is no longer “how do we manage better?” It’s “why manage at all?”

The signals are already visible. Klarna cut 700 roles, explaining that AI was doing the work. Development agencies are going from 40 developers to 5 developers plus agents. Consulting firms are halving their junior teams.

But these examples still operate within the old paradigm: keep the hierarchical structure, just swap some humans for agents in certain roles. It’s like electrifying a horse-drawn carriage. The real shift happens when someone realises they need neither the carriage nor the horse — just to get from point A to point B.

That’s what a singularity company is. Not a company that uses AI. A company that no longer needs to be a company in the organisational sense of the word.

---

The biological limit that AI is breaking

There’s a number in organisation theory called the “span of control” — the number of people a manager can supervise effectively. Studies put it somewhere between 3 and 8. Beyond that, supervision quality collapses. It’s a biological constant: the cognitive bandwidth of a human brain for managing coordination relationships is finite.

The entire architecture of modern companies flows from this constraint. If a manager handles 7 people, and you have 350 employees, you need 50 managers. To coordinate those 50 managers, you need 7 directors. To align those 7 directors, an executive committee. Every layer adds latency, distortion, and politics.

The org chart isn’t a design choice. It’s a cognitive prosthetic to compensate for the fact that humans don’t scale.

AI agents don’t have that limit.

One human can orchestrate 10 agents simultaneously without any loss of quality. Why? Because the nature of the relationship is fundamentally different:

• No emotional ambiguity. An agent doesn’t wonder whether your feedback was passive-aggressive.
• No information loss. Context is shared, versioned, persistent.
• No horizontal coordination needed. Agents don’t need to “align with each other” in the corridor.
• No calendar latency. No waiting until Monday for the team check-in.

The span of control goes from 7 to potentially unlimited. And when you remove the fundamental constraint that justifies hierarchy, hierarchy collapses — not because it’s bad, but because it’s unnecessary.

This is exactly what happened with telephone switchboards. For decades, every call required a human operator to physically connect two lines. Automating the routing didn’t “improve” the operator’s job. It made it redundant. Middle management is the telephone operator of 2026.

---

18 months solo + agents: what I’ve learned with no meetings and no managers

I’m not writing this from an ivory tower of theory. For eighteen months, I’ve been running an operation that generates the revenue of a small agency — alone, with an ecosystem of AI agents.

Here’s what a typical day looks like in a singularity company.

In the morning, I “manage” no one. I open my system, I check what my agents produced overnight — article drafts, data analyses, replies to prospects, code, monitoring reports. I validate, adjust, relaunch. In one hour, I’ve handled what would take a morning of meetings in a traditional structure.

There are no meetings. Zero. No daily standup. No weekly. No “quick five-minute catch-up” that runs 45 minutes. When I need to sync two workstreams, I do it in a prompt. Time required: 30 seconds.

There’s no HR process. No recruiting, no three-week onboarding, no annual reviews, no managing conflict between two colleagues who can’t stand each other. When an agent underperforms, I change its prompt or replace it. In five minutes.

There’s no organisational debt. This is probably the most underestimated point. In a traditional company, every hire adds relational complexity in a combinatorial way. Going from 5 to 6 people isn’t +20% complexity — it’s +40% relationships to manage. With agents, complexity is linear. Adding an agent means adding an agent. Full stop.

What have I lost? The serendipity of the coffee machine. Ideas that emerge from an unlikely conversation between two humans. That’s real, and I won’t minimise it. But let’s be honest: in most companies, the coffee machine mostly produces gossip, complaints about management, and weekend chat. Creative serendipity is the exception, not the rule.

What have I gained? Time. An obscene amount of time. And with that time, the ability to think deeply — which hierarchical management, paradoxically, prevents. Managers spend so much time coordinating that they have no time left to think.

---

The mycorrhizal network analogy: coordinating without controlling

To understand how a singularity company works, you need to abandon the org chart metaphor — that inverted tree with a boss at the top and executors at the bottom.

The best analogy I’ve found comes from biology: mycorrhizal networks.

Beneath every forest, there’s an underground network of fungi connecting the trees. This network transports nutrients, warning signals, information. When a tree is attacked by insects, the network transmits a chemical signal to neighbouring trees, which start producing defences before they’ve even been touched. When a young tree lacks light, larger trees send it carbon through the network.

There’s no boss. No hierarchy. No coordination meeting. The network operates through local signals and responses that produce coherent global behaviour.

A singularity company works exactly like that.

The human isn’t the “boss” of the agents. They are the initial signal — the intention, the direction, the taste, the judgement. The agents are the mycorrhizal network: they carry that intention, transform it into concrete actions, and send back signals (results, alerts, suggestions) that the human integrates to adjust the next signal.

This model has three properties that hierarchy doesn’t:

Resilience. If one agent malfunctions, the network keeps going. There’s no human single point of failure (except the orchestrator themselves — I’ll come back to that).

Speed. Information doesn’t travel through 4 hierarchical layers before reaching the person who can act. The intention → action → feedback cycle is near-instantaneous.

Adaptability. Reconfiguring the network takes minutes, not months. Need a new capability? Plug in a new agent. A market disappears? Unplug a workflow. No redundancy process, no reorganisation, no “change management” programme.

The limit of the mycorrhizal model is also its strength: it depends on the quality of the initial signal. A mycorrhizal network in toxic soil carries poison. A singularity company with a human giving vague directions produces noise at scale. The critical skill is no longer management — it’s clarity of intention.

---

What this concretely changes for anyone who wants to build alone

If you’re reading this and something resonates — if you’re a freelancer, solopreneur, independent, or an employee dreaming of building something — here’s what the singularity company model concretely changes.

The barrier to entry collapses. Launching an operation that produces at agency scale no longer requires raising funds to hire. The capital needed drops from hundreds of thousands of euros to a few hundred euros a month in AI tool subscriptions.

Competition changes in nature. You’re no longer competing with 50-person companies. You’re competing with other augmented individuals. The difference comes down to vision, taste, the ability to ask the right questions — not team size.

The winning profile changes. The “coordinator” manager loses their value. The profile that emerges is what I call the AI solopreneur: someone who combines deep domain expertise, the ability to orchestrate agents, and above all a sharp enough judgement to distinguish good from mediocre in what agents produce. That’s a rare profile today. It’ll be the norm tomorrow.

Legal structures need to evolve. A singularity company doesn’t need offices, collective agreements, works councils, or meal vouchers. Labour law and company law are built for human organisations. When the organisation is no longer human, these frameworks become empty shells — or worse, absurd obstacles.

Risk shifts. In a traditional company, risk is distributed: if the marketing director leaves, the company survives. In a singularity company, everything rests on a single human brain. That’s the model’s Achilles heel. The answer isn’t to reintroduce hierarchy — it’s to build agent systems that are documented and transferable enough to survive their orchestrator. We’re not there yet, but it’s an engineering problem, not a philosophical one.

---

The question that remains

I’m not claiming this model suits everything. Building a bridge, operating on a patient, teaching children — these activities will remain deeply human and collective.

But for everything that falls under intellectual production — writing, analysing, coding, designing, marketing, selling, advising — the singularity company isn’t one option among many. It’s an attractor that the market is converging towards, whether existing structures like it or not.

The real question isn’t “is this going to happen?” It’s already happening.

The real question is: which side of the transition will you be on in 12 months?

The one orchestrating the network, or the one discovering that their “team coordinator” role has just been replaced by a 200-word prompt?

---

Need a hand building something — a website, a SaaS, or an AI automation? Sébastien de Bollivier, the dev behind SEK, can help.

Also worth reading: the SEK journal.

Got an idea to ship? A website, a SaaS, an AI automation — built with you.

Talk about your project