Founders: How to Scale Your Organization Without Scaling Your Bureaucracy

At LinkedIn, I created a program called the Full Stack Builder (FSB) to combat a pattern playing out at nearly every company that scales. A builder's job starts out simple: have an idea, build it, and then take it to market. But, as the company grows, that simple process becomes very complex very quickly. Moments-in-time necessities create new layers: a large and complex research process, specific review cycles across design, trust, privacy, and security, and more checkpoints with each iteration. The actual "building" gets buried under coordination and what's left feels more like an assembly line.

To fix this at LinkedIn, we started to collapse the organizational stack, function by function, and rebuilt it around people who could own the whole loop themselves, from insight, to execution, to market.

‍“But as a founder, how do I scale without creating the very thing I eventually have to tear down?”‍

I was asked this question recently by a founder as part of my role as an advisor to Penny Jar Capital portfolio companies.

This is a unique challenge. It's one thing to tear down a machine you spent a decade building and rebuild it while it's still running. It's another to build one from scratch and get the blueprint right the first time.

You're already a full stack builder. Hire like one.

Here's what I'd tell any founder reading this: it doesn't matter if your own background is engineering, marketing, or design. By definition, you should already be what I'd call a full stack builder. A founder's job is a builder's job. You have an idea, and you take it to market yourself. Nobody hands you a spec.

But founders often leave that instinct behind when they start building a team. They hire a specialist for niche parts of the work, and without meaning to, they rebuild the exact bureaucracy they tried to avoid. Instead, they should think about building their team with full stack builders from the get-go.

Two Must-Haves As You Scale

First, I would look for AI agency: the ability to use the tools available right now to move faster with higher quality, without waiting to be handed a playbook. AI-native talent has already torn apart and rebuilt how they research, prototype, and ship around the tools that exist today, and they'll keep doing it as the tools change under them.

Second, I would hire for judgment. If you play the long arc forward, a lot of the execution work you'd hire a specialist for today will be solved by AI tools within the next model generation or two. But judgment, the ability to make high-quality decisions in complex and ambiguous situations, won't be replaced by AI tools (at least not any time soon).

AI Fluency is not AI Agency

If I were hiring someone today and wanted a real proxy for their ability to excel regardless of how far models and agentic workflows evolve, here's what I'd bet on: their ability to think critically about AI output instead of just accepting it; to plan before they execute; and to exercise taste about what to build, and when not to build. None of that comes from a model; it comes from them.

To uncover AI agency, the interview itself has to change. At LinkedIn, we built the Associate Product Builder Program specifically to test for it, a two-year rotation for entry-level talent that skips the resume entirely. Candidates submitted a product they'd built instead. The final interview wasn't a conversation about how they'd approach a problem. It was building something live, end-to-end, in the room: speccing, designing, prototyping, shipping, all of it, in real time.

Watching those interviews taught me that the gap between AI fluency and AI agency is bigger than what people think. Some candidates could talk fluently about AI research, cite the latest release, and sound sharp at a coffee conversation, but didn't build anything of substance with AI. They still worked the same way. The ones who stood out rebuilt their own workflow around the tools: they moved between them quickly, shaped input into output, and knew exactly when to trust what came back and when to push on it.

Lessons Learned Signal Judgment

Judgment is different. When I first laid out the FSB model, I put five human traits at the center of it: vision, empathy, communication, creativity, and judgment. All five still matter, but judgment is the one that decides whether the other four, and all that agency, add up to something good or just something fast.

Judgment comes from the ability to learn, lived experience and real lessons learned: scaling something, running into a genuinely gnarly problem, and knowing how to work through it. That doesn't mean the person has to be senior in years. It means they've been tested, learned, and they know what dealing with a hard problem actually feels like in their hands.

The takeaway for founders isn't that every hire needs to max out on both. It's that AI agency and judgment both need to show up across the team, in different degrees from person to person. You can test agency directly: hand someone a real problem, let them use whatever tools they want, and watch what they build. Judgment shows up differently. Probe the lessons they've learned: what actually went wrong the last time they hit a hard problem, what they tried, and what they'd do differently now.

Build a team of full stack builders who can own the whole loop. Test for AI agency. Probe for judgment. That's how you scale the company without scaling the bureaucracy.