The Org Chart is a byproduct of Roman wars

AI slowly eats away the modern org chart

Did you know that the modern org chart is actually modelled after how soldiers reported to generals during war times in the Roman period? Just like how schools are modelled after the Germanic assembly line (aka outdated), org chart make-ups isn’t that innovative either. And it’s controlled us for about 2000 years. Let that sink in.

So apparently, back in the day, 8 soldiers would form a contubernium (pictured below)

Which really was just a shared leather tent for the eight dudes. They shared a mule and equipment, cooked together and carried gear together and were responsible for each other’s discipline.

And then if we look at a whole army:

• 8 soldiers → contubernium (led by a decanus)

• 10 contubernia → 80 soldiers → a century (led by a centurion, yes that’s the name of the Amex lounge at all the airports!)

• 6 centuries → 480 soldiers → a cohort

• 10 cohorts → about 4,800–5,000 soldiers → a legion commanded by a legate

And that’s how Alexander the Great conquered Europe.

Just kidding, he was not Roman.

I just couldn’t recall a super famous Roman general off the top my head.

But you get the point.

It was this way of structure that was passed down for centuries, where now we have the Individual Contributor (IC) reporting into a manager, who then reports into a Director, VP etc etc. But when AI comes into the mix, this org chart idea might become a thing of the past. That certainly is the case at many tech companies these days.

In a recent essay Block CEO wrote called from Hierarchy to Intelligence, he talks about how he will be facilitating the re-org of the company.

Block: Three People Doing the Work of Fourteen

Block is a company a lot closer to you than you think. They own Square. Which is what you approach every time you’re paying for that damn over inflated $8 latte. And it always has the nerve for you to choose how much to tip on top of the $8 coffee. Okay sorry for the rant.

It’s the same company that announced it was cutting more than 4,000 employees, roughly 40% of its workforce a few weeks ago. They let people go not because the company was doing poorly financially. In fact, it was growing 17% YoY. Instead, he thinks that corporate hierarchy has always existed to solve one problem: routing information through organizations too large for any single person to oversee. AI can now perform that function continuously and at scale, making the human messenger layer redundant.

In the future, his company will have three types of roles:

·  Individual Contributors: deep specialists who receive context from the AI system (not a manager), allowing them to make decisions without waiting to be told what to do.

·  Directly Responsible Individuals (DRIs): people who own specific cross-cutting problems for defined periods (90-day cycles) and have full permission to pull whatever resources they need from the system.

·  Player-Coaches: the replacement for managers. They both contribute individually (still building) and develop talent. As Dorsey puts it, there is no need for a permanent middle management layer; player-coaches combine craft and people.

So is the DRI the old SVP or PM?

In Dorsey’s world, the DRI is not a rank, it is a temporary assignment. That is the key difference. An SVP holds a permanent position in the hierarchy with a standing team, a standing budget, and standing authority. A DRI owns a specific outcome for a defined period, typically 90 days, and then the assignment rotates or ends.

The idea is that instead of having a senior vice president who permanently owns "payments infrastructure" with six directors and forty engineers reporting up through them, you have a DRI who owns "solve the merchant cash flow gap problem" for one quarter, with full permission to pull whatever resources they need from across the company. When the problem is solved or the cycle ends, that person either picks up a new problem or goes back to building.

In theory, this eliminates the empire-building that happens in traditional structures, where SVPs accumulate headcount, protect territory, and create bureaucratic drag. The DRI model keeps authority fluid rather than structural.

From Bees to Beekeepers

In my Hive Structure framework, the distinction between bees and beekeepers has never been more concrete than it is in Block's new model. The bees are no longer human. The bees are the AI systems doing the volume work at machine speed.

The three human roles that remain are all beekeeper roles.

The individual contributor is the beekeeper who holds deep domain expertise that the AI cannot replicate, the person who knows why a particular architectural decision matters or where a regulatory line sits. The DRI is the beekeeper who decides which flowers the hive should fly toward this quarter, setting direction and owning outcomes that require human judgment about tradeoffs the system cannot weigh on its own. The player-coach is the beekeeper who tends to the other beekeepers, developing talent and maintaining the culture and craft standards that no model can set. None of these roles exist to do the volume work. They exist to direct it, judge it, and ensure it serves the right purpose. That is what it means to move from bee to beekeeper: you stop producing the honey and start deciding what the hive builds next.

Meta: The 50-to-1 Experiment

If Block is the company that published the manifesto, Meta is the company stress-testing the most extreme version of the thesis at scale.

Meta's new applied AI engineering organization operates at a 50-to-1 employee-to-manager ratio, double the outer limit most software organizations consider functional. Internal AI agents now handle the administrative and project management tasks that once required thousands of middle managers: tracking projects, identifying blockers, assigning work, and sharing critical information faster than any human reporting chain.

The restructuring goes deeper than ratio changes. Meta has established aggressive internal targets for 2026, requiring engineers to use AI for up to 75% of their coding tasks. Teams are being reorganized into smaller "AI Pods." Engineers are being rebranded as "AI Builders." An internal gamified system called "Level Up" rewards employees with badges for meeting AI adoption milestones. An "AI Performance Assistant" helps managers write reviews, automating yet another layer of what middle management used to do.

For Meta, this is more of a money move. Meta is spending up to $135 billion on AI capital expenditure in 2026, nearly double its 2025 outlay. CEO Mark Zuckerberg has publicly committed to over $600 billion in U.S. AI infrastructure investment by 2028. That money has to come from somewhere. Reports indicate that senior executives have been directed to plan workforce reductions of 20% or more, potentially eliminating roughly 15,800 of approximately 79,000 jobs. Wall Street approved: the stock rose roughly 3% on the news, with analysts estimating a 20% cut could add about 5% to earnings. Cha ching! More money to build data centers.

The Uncomfortable Truth About Career Ladders

We cannot discuss flattening org charts without acknowledging what it means for the people inside them. The org chart does not just encode reporting relationships. It encodes compensation, status, and career progression. Telling a generation of middle managers that their career ladder just lost three rungs is not merely a structural exercise. It is a deeply political one.

The companies that navigate this transition well will redefine career progression around depth of judgment, scope of AI system design, and the ability to operate at the lateral intersections that create the most value. The companies that do not will either bloat with unnecessary layers to preserve egos, or flatten too aggressively and lose the institutional knowledge that those middle layers were quietly holding.

There’s just one more problem

The vision is elegant: replace the hierarchy with an intelligence layer that gives every person real-time visibility into the entire company. But the infrastructure to do that does not exist yet. At most companies, context still lives in fragments across Slack, Jira, email, code repos, and Notion. The hardest part of the whole thesis is not cutting managers or renaming roles. It is building the connective tissue that lets a DRI actually see what a hundred other autonomous teams are doing without picking up the phone. Until that layer exists, you have a hundred strike teams operating with executive authority and limited visibility, which is not a flat organization. It is organizational fog of war.

If you want to read more about org charts and the possible changes it brings to the way we look at work, check out my book How to Do More with Less Using AI (Wiley)as it’s  finally released!

Sharon Gai is an AI transformation strategist, keynote speaker, and author of How to Do More with Less Using AI (Wiley). For speaking inquiries, visit sharongai.com.