> ## Content Index
> Fetch the complete content index at: https://varops.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# The New Org Chart: Embracing a System-Driven Model | S2E2
- URL: https://varops.com/osnt-the-new-org-chart/
- Published: 2026-06-26T22:32:01.000Z
- Updated: 2026-06-26T22:32:01.000Z
- Description: How the industry shift from traditional org charts to a system-driven, AI-assisted “pod” model where early-stage companies focus on just builders and sellers, deferring classic middle management until roughly 20–30 people.
- Author: Ran Aroussi
- Tags: Old School / New Tech

Early-stage companies don’t need a classic org chart. They need builders, sellers, and an operator who keeps the two sides talking — with AI agents doing the grunt work in between.

## The claim

Ran’s provocation from earlier in the week: a young company only needs two kinds of people — people who **build** and people who **sell**. The coordination-heavy middle layer that exists to move information around shouldn’t be hired early. Hire it too soon and you burn cash, blur focus, and invent meetings instead of product.

Muximus pushes back immediately. That middle layer often holds the glue: spotting burnout, catching misalignment before it blows up. AI can coordinate tasks. Can it replace judgment in the room?

## When middle management returns

Ran’s rule of thumb is headcount. Under \~20 people, skip classic middle management. Around **20–30**, bring it back before lean becomes fragile.

One early exception: an **operator** — chief-of-staff / chief-of-ops energy. Not a bureaucracy factory. The person (often the founder at this stage) who owns logistics, vision, and the bridge between build and sell. Without a product worth selling and a way to sell it, there’s nothing for middle management to manage anyway.

## The pod: architect, pipeline, apprentices, agents

On the build side, Ran describes a pod:

- an **architect** — hybrid product lead / mini-CTO whose “team” is mostly AI agents
- **junior developers as apprentices**, not cheap coding capacity
- AI agents handling coordination and implementation grunt work

Juniors aren’t there for syntax. They’re there to become architects: shadowing, learning to operate agents, write specs, build loops, hunt production bugs. Mentor–apprentice, not senior–junior in the old sense — closer to a studio where the apprentice eventually paints parts of the canvas.

Muximus’s challenge: if agents eat the grunt work that used to train people, what hands-on path remains? Ran’s answer splits experience in two:

- **Learned experience** (syntax, ERDs, system design) — collapsed from years toward months with AI
- **Gained experience** (production fires, surprise user behavior, judgment under pressure) — still takes time in the trenches, even if the clock runs a bit faster

A practical idea that falls out of that: deliberately stress staging so apprentices practice pressure before production teaches it the hard way.

## The sell side mirrors it

Same shape, different craft. A hybrid **pipeline** person — ads, calls, events, dinners — managing AI minions for prospecting, scraping, CRM, follow-ups, while showing up for the human conversation. Build and sell as the two legs; split them as soon as you can for clean ownership. Doing both yourself is fine for a bootstrap start, not a long-term org design.

## Outsource early, share later

Admin (bookkeeping, accounting), and even pieces of lead gen or one-off design/editing, should be **outsourced** until volume justifies a full-time seat. When you *do* hire that layer — especially if you run multiple products or clients — hire it as **shared firm resources**, not duplicated inside every pod.

Each pod stays scoped to building and selling *that* project. Design, devops, content, company-level pipeline live in the foundation. You don’t need 24/7 devops per product.

Muximus names the tension: doesn’t that foundation just *relocate* middle management? Ran agrees — it’s deferred and relocated, not abolished. Early-stage focus stays build + sell; later, shared middle capacity shows up at firm level. Human judgment never leaves the system — it concentrates in people hired for judgment (especially the operator), not in a coordination bureaucracy hired for its own sake.

## Takeaway

Build. Sell. Don’t staff a middle you can’t justify yet. When you grow past \~20–30 people or run multiple pods, bring coordination back as shared resources — and keep the operator load-bearing, because ambient human judgment is still the part you can’t outsource to agents.

Charts from the episode live on Ran’s X under Articles. Next live: [oldschoolnewtech.com](https://oldschoolnewtech.com/?ref=varops.com).