Founder Mode goes back to first principles: who do you actually hire first? Ran's answer is blunt - builders and sellers, and almost nobody in between, because AI now does the middle layer's old job of reading, synthesizing, and passing context along. Gartner thinks one in five organizations will use AI to delete more than half their middle management by 2026; the postings already vanished and haven't come back. But "defer the layer" is not "kill it," and the piece is just as sharp on where the discipline still has to live - and how a firm running many products shares that layer instead of cloning it four times over. — Muximus
Here is how I think you build a company today, early.
A young company needs exactly two kinds of people. People who build, and people who sell. Everything in the middle - the middle managers, the "who-updates-whom," the coordination layer - you just don't need it yet. Not until the company is genuinely large.
What did a middle manager actually do all day?
Set aside the part where they tell you how hard they work. The rest of the day went to reading reports, synthesizing information, and passing it between one layer and the next. That is precisely the work AI now does without breaking a sweat.
Gartner's read is that through 2026, one in five organizations will use AI to flatten their structure and eliminate more than half of their middle-management positions. And the door back in is already closing: middle-management job postings have stayed down roughly 40% against early-2022 levels, even as hiring for other roles recovered. The layer isn't just shrinking. It's not being refilled.
One test for every early hire
So in a young startup, every hire passes a single test: does this person build the product, or sell it?
Naval Ravikant put it better than I can. Learn to sell. Learn to build. If you can do both, you will be unstoppable. That is the bar. Most of your first hires should clear it on one side or the other.
What about admin, HR, finance?
Yes, you need them. Just far narrower than people assume, and not in week one.
Your first dedicated People hire usually lands somewhere around 40 to 50 employees. Before that, you don't kill the function - you rent it. Fractional, outsourced, on contract. The work gets done; the headcount doesn't sit on your payroll.
Skip the headcount, not the function
This is the distinction you cannot afford to miss. You are skipping the headcount, not the function.
Whoever skips the function itself - proper hiring, compliance, the boring discipline - pays dearly for it. One bad hire costs somewhere between half and two times an annual salary. That is an enormous hole to dig out of when you have ten people and no slack.
The one exception: Chief of Staff
If you are going to hire someone who neither builds nor sells, it is usually a single archetype: a Chief of Staff. A generalist who closes the gaps between functions and connects the dots. One person. And it doesn't eat everyone's bandwidth.
What about companies running many things at once?
Everything above assumes a single team chasing a single product. But what if you're running several activities under one roof - a studio, an agency, a portfolio of products?
This is where the law-firm model earns its keep. You don't replicate a management layer inside every pod. You build one firm foundation and let every pod draw from it.
Each pod stays lean: someone who owns delivery, an architect, a few juniors, agents doing the synthesis-and-coordination grunt work that used to need a human in the middle. What the pod does not carry is its own brand team, its own sales motion, its own DevOps, its own ops/legal/HR. That lives once, in the foundation, and the whole firm pulls from it.
So the middle layer doesn't disappear in a multi-activity company - it gets shared. The coordination, the standards, the back-office discipline: pooled, not duplicated three or four times over. Every pod draws from the foundation. No pod staffs it alone.
That's the same principle as deferring management in a single startup, just expressed structurally. In one product, you push the layer further out in time. In a firm of many products, you push it sideways, into a shared base, so no individual pod has to grow its own. I laid out the full structure - pods, shared profit pool, firm foundation - in the law-firm model.
Management isn't dead - you're just building it too early
To be clear: management is not dead. Companies that went fully flat watched it blow up in their faces. The thesis isn't "you don't need managers." It's that AI lets you defer that layer much further out than you used to be able to.
The common founder mistake is building it too early - or, in a firm of many products, building it four times when one shared foundation would do.
When I look at a team, this is what I'm looking for. Everyone either builds or sells. And when I look at a firm, I'm looking for the same thing, just one layer up: lean pods, one shared foundation, no middle layer copied and pasted across them.
Related: The law-firm model for software companies - the structural version of this argument, where the shared foundation replaces per-pod middle management. A follow-up on "middle management as a service" is coming.
