How Organized Effort Looks Like Post AI

Everyone is talking about what AI builds. Almost no one is talking about how we organize to build it.

This is the bigger problem. And almost nobody sees it yet.

I’ve spent time in industrial automation. I watched factories go from hundreds of workers on a line to a handful of operators managing robotic systems. The pattern was always the same. First, the machines replaced individual tasks. Then they replaced entire processes. Then someone looked around and realized the org chart made no sense anymore. All those layers of supervision, all those coordination roles, all those managers managing managers - they existed to solve a problem that the machines had already solved.

The factories that adapted fastest didn’t just buy better robots. They redesigned how humans were organized around those robots. The ones that didn’t adapt kept their old structures, added machines on top, and created expensive chaos.

That exact pattern is now playing out in knowledge work. Except it’s happening in years, not decades.

---

The execution problem is solved

AI can write code. It can draft legal documents. It can build financial models, create marketing campaigns, generate research syntheses, and produce first drafts of almost anything a knowledge worker gets paid to do. GitHub reports developers merged 43 million pull requests per month in 2025, up 23% from the prior year. A quarter of current Y Combinator startups have 95% of their code written by AI. Companies are hitting $10M in revenue with teams under 10 people.

The execution bottleneck is gone. Or at least, it’s collapsing so fast that planning around it is a mistake.

Brynjolfsson’s research showed a 14% average productivity increase in call centers using AI, with 34% improvement for novice workers. Harvard found a 40% time reduction in writing tasks. These aren’t projections. These are measured results from the last two years. And the tools are getting better every quarter.

So what’s the new bottleneck?

Coordination. Judgment. Accountability. Deciding what to build, why to build it, whether it’s working, and who is responsible when it isn’t.

These are organizational problems. And most organizations are still structured to solve an execution problem that no longer exists.

---

The org chart is a fossil

Think about what a traditional company structure actually is. It’s a system designed to coordinate human labor at scale.

You have departments because specialization made individual humans more productive. You have middle management because information needed to flow up and decisions needed to flow down. You have hierarchies because someone had to ensure that hundreds or thousands of people were pulling in roughly the same direction.

Every layer exists to solve a coordination challenge created by having lots of humans doing lots of tasks.

Now remove the need for most of those humans to do most of those tasks. What happens to the layers?

They become friction. They slow things down. They add cost without adding judgment. They create meetings about meetings. They diffuse responsibility until nobody owns anything.

I’ve seen this in real organizations. A team of 30 people where maybe 5 are doing the actual work and 25 are coordinating, reviewing, approving, and reporting on that work. Remove the execution bottleneck with AI and what do you have? Five people orchestrating AI systems and 25 people with nothing to coordinate. That’s not a restructuring problem. That’s an existential one.

Shopify’s CEO made headlines in 2025 by telling managers they had to prove AI couldn’t do a job before hiring a human. That got attention because it sounds bold. But it’s actually a half-measure. It optimizes headcount within the old structure. It doesn’t question whether the structure itself still makes sense.

And it doesn’t.

---

What actually needs to change

Most people think the shift is from “big teams” to “small teams.” That misses the point entirely. The real shift is from organizations designed around executing tasks to organizations designed around exercising judgment.

When AI handles execution, the humans who remain need to do the things AI cannot: set direction, make judgment calls under uncertainty, maintain accountability, build trust-based relationships, and keep the whole system honest.

These functions don’t need departments. They don’t need middle management. They don’t need org charts with dotted lines and matrix reporting. They need clear ownership. One person, one function, full accountability.

My strong opinion after years of thinking about this - studying the research, building organizations, watching what works and what doesn’t - is that the company of the future has a specific shape. It’s small. It’s flat. Every human in it holds real authority over a defined domain. And it scales not by adding people, but by adding AI leverage underneath those people.

I call this model the Cell.

I’m not going to fully explain it here. That’s coming in the next article. But I want to lay out the principles that any durable post-AI organizational model must satisfy, because these are non-negotiable regardless of what you call the structure.

Individual accountability, not shared responsibility. Every critical function needs one person who owns it completely. When everyone is responsible, no one is responsible. This is true in a team of 5 and true in a company of 5,000. AI makes it more urgent because the speed of execution means mistakes compound faster than ever.

Human judgment at the top of every decision chain. AI executes. Humans decide. Not because AI can’t make decisions - it increasingly can - but because someone has to be accountable for what those decisions produce. Legal liability, ethical responsibility, and stakeholder trust all require a human at the top.

Small enough for trust, large enough for challenge. A single person can’t challenge their own thinking. A team of fifty can’t maintain the trust and speed required. The sweet spot is somewhere in between - small enough that everyone knows everyone, large enough that genuine disagreement is possible.

Generalists over specialists. When AI can provide specialist knowledge on demand, the premium shifts to people who can synthesize across domains. The most valuable humans in the AI era are the ones who understand enough about everything to direct AI systems intelligently and catch their failures.

Designed for leverage, not headcount. The measure of organizational capacity is no longer how many people you have. It’s how much each person can direct. One skilled human orchestrating AI agents and automated systems can produce what a traditional team of 10 or 20 produced. The organization should be designed around that reality.

---

The timeline

I want to be honest about what I know and what I don’t.

I believe these principles are valid for the period between now and roughly 2040. That’s the window where AI capabilities are growing fast enough to transform knowledge work but haven’t yet reached the point where they can fully substitute for human judgment across all domains.

What happens after 2040 depends on questions nobody can answer yet. If artificial general intelligence arrives on the timelines some researchers predict - and those timelines keep getting shorter - then even these organizational principles may need rethinking. The median expert prediction for AGI has compressed from 50 years in 2020 to roughly 5-7 years on current forecasting platforms. That should make everyone uncomfortable, including me.

But here’s what I know from watching automation transform industries for years: the organizations that fail are never the ones that couldn’t adopt the new technology. They’re the ones that couldn’t reorganize around it. The technology is the easy part. The human structure is the hard part.

Most companies right now are bolting AI onto their existing structures and hoping for the best. McKinsey found that 95% of generative AI pilots fail to deliver meaningful business impact. That number should terrify executives, but it shouldn’t surprise anyone. You can’t solve an organizational problem with a technology purchase. The AI works fine. The structure around it was built for a different world.

---

What comes next

This is the first article in the Company of the Future series.

In my Jobs of the Future series, I mapped five roles that remain fundamentally human as AI transforms work. In my Skills of the Future series, I explored the capabilities those roles demand. This series completes the picture: how do you actually organize those people into a structure that works?

Over the coming weeks, I’ll go deep on the organizational model I’ve been developing. I’ll start with the Cell - the core unit of the post-AI organization. Then I’ll show how it works at different scales: the single operator, the triad, the full five-person Cell. And I’ll address the hardest question of all - how existing organizations can evolve toward this model without burning everything down.

This is not theory. This is a framework built from real experience designing organizations in high-complexity environments, validated against the best available research, and stress-tested against the reality of how AI is actually changing work right now.

The product problem is solved. AI builds things. The organization problem is wide open.

That’s where the real work begins.

---

This is Part 1 of the Company of the Future series. Next: the Cell - the smallest viable unit of human organization in the AI age.