Save a Seat for AI: Notes on Becoming an AI Native Company

Back in late 2022, Elon Musk took over Twitter and cut about 80% of the staff — from roughly 8,000 down to 1,500. Most people thought X would be a smoking crater within months. Four years later, it's still there, still functioning, arguably thriving. Make of that what you will.

Lately, the story has been rhyming. In February, Block announced layoffs of about 4,000 people — roughly 40% of the company. In late April, Meta said it would let go of around 8,000 more, about 10% of headcount, and started executing in May. The contexts and reasons are different, but there's a common thread underneath: the org chart everyone grew up with is being seriously rethought.

From Hierarchy to Intelligence

Not long ago, Jack Dorsey and Sequoia's Roelof Botha co-wrote an essay called From Hierarchy to Intelligence, framing the Block layoffs not as cost-cutting, but as a permanent organizational restructuring.

Their argument: corporate hierarchy was essentially copied from the Roman military's information routing system about two thousand years ago. The role of a middle manager, stripped down to its core, is to be an information relay — pass it up, pass it down, summarize, translate. And that is exactly the job AI is now competent enough to do.

I have my reservations about some of their specific prescriptions. People are creatures of habit. Telling someone whose role has been fixed for 20 years to suddenly "operate differently" usually doesn't produce efficiency — it produces ambiguity and finger-pointing. Re-orgs are hard. Re-orgs that try to flatten the entire structure overnight? Even harder.

But the underlying idea? That part I agree with completely:

Companies need to start treating AI as an important colleague, not a "tool."

That's also the direction YC has been hammering on recently with the AI Native Company concept.

Personal AI vs. Company AI

If you've been using AI seriously for a while, you've built up a kit of personal tricks. Which model is good for what. How to format docs so AI actually reads them. How to phrase a prompt to get something usable on the first shot. Every conversation with another power user is a chance to swap a new trick.

But the AI agent in my head isn't just a personal copilot. It's something that can access the entire company's data, history, and context. What changes when that's true? How does it actually accelerate growth?

A few things I've been chewing on.

1. Save a Seat for AI

The capabilities of AI agents are going to keep evolving. We can't define their role inside the company based on today's ceiling — by the time you finish writing the job description, the ceiling has moved.

But the opening move is simple: give AI a friendly environment to gather context first. Don't rush to assign it a title or a swim lane. Just save it a seat at the table, and start feeding it everything you'd feed a smart new hire.

2. Everything Is Queryable

This is the line YC keeps repeating, and it's the right one. Every piece of information — decisions, data, documents, meeting notes, customer feedback — should be searchable and accessible by AI.

More context, better decisions. And unlike humans, AI doesn't clock out at 6pm.

The hardest part of this isn't technical. It's cultural. It means writing things down when previously you wouldn't have bothered. It means choosing systems that expose their data instead of locking it up. It means treating documentation less as overhead, and more as feeding your most diligent colleague.

This is also why I've been so deep on the Business as Code idea — once your business lives in plain text in a Git repo, "queryable" stops being a roadmap item and becomes a default.

3. AI Is an Amplifier

When you amplify every individual's capability, the lines between roles get blurry. Fast.

Engineers can do PM-ish work. PMs can ship a bit of code themselves. Marketing can solve problems they used to file tickets for. The old "I need to find someone who can do X" workflow gets compressed dramatically.

A startup that used to need 10 people and two years to ship a product can now do a rough version with two people in a week. Companies need fewer people overall (which, yes, is a big part of what these recent layoff waves are really about) — but each person can accomplish much more.

The result: role definitions become goal-oriented instead of function-oriented. Decisions stop getting stuck in hierarchy because every individual can pull the context they need through AI and contribute directly. The prerequisite is that information has to be more transparent, more objective. Politics dies a quiet death when the data is right there.

4. Close the Feedback Loop

AI itself evolves through a loop: Input → Output → Eval → Input. 24/7, no breaks.

The question is whether your organization runs on the same loop. Are your information flows, your people, and your collective intelligence all feeding back into something that learns? Or are decisions and outcomes disappearing into Slack threads and shared drives, never to be evaluated again?

The companies that figure out how to close this loop — at the org level, not just the personal one — are going to compound faster than everyone else. It's the difference between a company that gets smarter every quarter and one that keeps re-learning the same lessons every time someone new joins.

5. The Talent Layout Question

Here's the part fewer people are talking about.

In the AI Native Company frame, costs are lower and revenue has to come faster. The obvious move is to shrink headcount. The less obvious move is to rethink what kind of people you want in the building at all.

When every individual's capability is dramatically amplified, a lot of them will naturally want to start something — build a tool, sell a thing, run a side project. Trying to suppress that instinct is a losing battle.

So instead of forcing people into traditional roles or watching them leave to do their own thing in isolation, what if the company became a platform? An umbrella that lets people use company resources to collaborate on projects, with shared upside.

That's what Individual Contributor could actually mean. Not a level on a career ladder. A real, contracted individual operator who works on different projects under the company's umbrella. The company becomes more like a matching platform that ties different ICs to different products, and lets everyone's incentives align.

Why This Is Hard to Talk About

These ideas mostly crystallized while I was watching a bunch of YC videos on the AI Native Company concept. My own company is moving in this direction, carefully.

The hard part isn't doing it. The hard part is that the concept is still new enough that explaining it to anyone who hasn't already been thinking about it doesn't quite land. You describe a flat-ish, AI-augmented, IC-platform company and people nod politely while clearly imagining you're describing either a glorified gig economy or a startup that hasn't figured out management yet.

Neither, really.

It's something different. And I suspect we'll have a much better vocabulary for it in about two years — once the early movers have visibly compounded ahead.

For now, the playbook is simpler than the theory: save a seat for AI. Make everything queryable. Treat people as amplified. Close the loop. And start letting "individual contributor" mean what the words actually say.