Growth Academy · Section 9 · Running the machine
If you disappeared for two weeks, what in your company would actually stop?
Nobody sets out to build a company that runs through one head. It happens by accident, and it happens for a good reason: someone is competent, someone is available, and every hard call routes to them because routing it there is the path of least resistance today. Do that for a year and the person who was simply good at their job has quietly become a single point of failureAny one person, tool, or channel whose loss stops the whole business. for the whole business.
Picture a lead developer who is coding and writing detailed instructions for every other developer at the same time, with no room left to also step back and manage. The right diagnosis isn't a workload problem. It's a structural one: power and information have centralized on one person, and the fix isn't "work harder." It's finding where the centralization lives and deliberately taking pieces of it away.
That's the whole section. Not "hire more people" and not "trust your team" as a slogan. Find the concentration. Decide, item by item, whether the fix is a documented structure anyone can run, a person coached into owning it, or information that needs to leave one head and enter a shared system. Then hold the culture that makes people willing to be coached, and willing to hand things off, in writing, so it survives past the moment you happened to say it out loud.
Start with the uncomfortable diagnostic question: for every recurring decision, approval, account, password, and relationship in your company, who does it currently route through? If the honest answer for a growing list of items is "one person," you have already found your risk. The person is usually good at the job. That is exactly why nobody noticed the concentration building.
"One of the key things our job as founders is to recognize, where is the power being centralized? The way we do it is with first, with understanding of where the mechanics are … dissecting his work, sitting down, understanding what he's doing, what he's not doing, and taking away some of the things from him."
Decentralizing isn't a declaration. You don't get to announce "we're decentralized now" and have it be true. It's dissection: sit down, understand what the bottleneckThe one point where work piles up because everything has to pass through it. person actually does day to day, what they don't do, and pull specific pieces away, one at a time, by giving other people enough information, tickets, stories, clear designs, that they can own the piece without routing back through the same door.
Power isn't the only thing that centralizes. Information does too, and it's the quieter version of the same failure. A founder who writes a detailed weekly status report can broadcast it and still keep the company dependent on their read of it, unless the team is actively trained to question it back: which screens got done, is there a link, what does this number actually mean. That training goes both ways. It makes everyone responsible for asking, including the person who wrote the report in the first place.
The same risk shows up one level up, in whatever lives only in the founder's own head. One founder pushed every co-founder into fundraising-focused rooms even while running the raise personally, because investors pull every founder into diligence regardless of who normally owns it, and because the business has to survive if the person who knows the most becomes unavailable. "What happens if I'm sick? What happens if a car hits me tomorrow? Are you going to let all of our work go to waste? You guys need to rise up." That's not a hypothetical founders enjoy sitting with. It's the actual test.
None of this works if the people you're handing pieces to don't actually own the plan. After walking a founding team through a set of ranked decisions, the wrong question is "does everyone agree?" Agreement is passive, and it's reversible the moment things get hard. The right question is what each person individually needs changed, added, or removed to own the plan as if they'd written it themselves. Ownership is what survives. Agreement is what quietly turns into "I just followed your vision" six months later, when there's no founding team left, just a CEO with reports.
Define your role by the areas where you create outsized impact, strategy, product, marketing and sales, rather than by a list of tasks you personally execute. That reframe does something practical: it also forces you to name, early and out loud, what you will not do. One founder named fundraising, firefighting, and day-to-day people management as areas he wasn't taking on, so the rest of the founding team could plan around a real constraint instead of an assumed availability.
The second half of role design is matching enjoyment to skill, not just skill to task. Everyone can tolerate some amount of work they don't enjoy. The problem is what happens to the tasks that fall outside that overlap: they get filed into a mental bucket, and as the bucket fills, the person starts looking for an exit. The design goal for a founding team isn't competence matching alone. It's maximizing where "good at" and "wants to do" overlap for each person, and treating the mismatch bucket as an early warning, not a personal failing.
Even that isn't a fixed org chart. Generic frameworks, a CTO manages, a CTO doesn't write code, describe a mature-company default, not a rule that applies at every stage. Early on, who does what should be decided by where responsibility and passion genuinely overlap, and renegotiated the moment a specific task becomes a bottleneckThe one point where work piles up because everything has to pass through it., not imposed from a template. And a flat, multidisciplinary structure is the right model when every contributor is a genuine domain expert: like a band where two names get the songwriting credit but every other member's part was real and necessary. The catch is external, not internal: a team can be flat inside and still needs a name outside who's accountable, or it reads as leaderless, a company in search of a leader.
Once roles are set by impact and overlap, there's still a separate question: what actually makes a given person effective in their role. The answer is three components, and having only one or two of them isn't enough.
The hierarchy
Who reports to whom, how often they plan together, how their work gets evaluated.
The structure, communicated
Whether that structure is actually written down and stated, not just implied. Is there a scheduled meeting? Is the reporting line explicit?
The personal touch-point
A separate outlet from planning and review. You cannot push or influence someone you have no relationship with, no matter how clear the structure is.
The relationship leg has a specific, small shape: a dedicated touch-point meeting, five to twenty minutes, distinct from planning ("what are we doing") and review ("did we deliver, why or why not"). It's a coaching conversation where the leader mostly listens: how are you doing, what's going on, how do you think you did, what could be better. The cadenceA fixed rhythm of meetings and check-ins the business runs on, such as a weekly review., how often it happens, scales with impact: monthly at minimum for anyone, weekly for the people carrying the most.
Handing someone a leadership title before they've earned trust and gotten real mentorship doesn't produce leadership. It produces confusion. Picture a technical lead given ownership of a workstream with no founder-level direction behind it: "leadership doesn't just happen by giving them the ability to lead … they need mentorship to lead." The fix wasn't taking the ownership back permanently. It was leading closely and explicitly for a defined window, then handing real autonomy back once the person had shown they could deliver under structure.
The same restraint applies to anyone giving feedback to someone with real gaps. If you can see everything wrong with how a person is working, unloading all of it at once overwhelms them, and the likely outcome is that they quit or shut down. Coach the way you'd coach a hockey player: pick the single most important thing right now, get a small, visible win, let trust build from the progress, then expand into the next area. The visible progress is what earns you the right to ask for more later.
Coaching also has a hard limit worth naming plainly: you cannot force someone to bring their best. The only lever that reliably works is building an environment people actually want to be part of, one where a real challenge gets brought forward and trusted to be solved together, instead of hidden or panicked over. And when you want to know what that environment is actually missing, the sharpest question isn't open-ended. Ask each person, one at a time: if you had a magic wand and could change one thing that would make the biggest difference, what would it be? Pair it with explicit permission to be honest, this recording is only for us, if you want to say something you don't want to say, I can pause it. A single forced priority beats a diplomatic list every time.
Coaching and redistribution only stay safe if the culture underneath them is explicit, not assumed, and one distinction does most of the work: a value is something you can't switch off, "you are either this or not," where a working style is just a preference. Teamwork can be a preference. Integrity, defined operationally in a written culture document, is a value: always doing what you said, the way it was meant to be done, by the time it was meant to be done, within the agreed conditions. The felt signal of a breach is that something feels incomplete, work that's technically done but not done the way it was meant to be.
That value, and the ones built on it, only function in a specific order. Skip the order and a demanding culture quietly turns into an excuse-making one.
"This order is not haphazard. This order builds on each other. If any of them fail … you stop here and you deal with that. Then you move to the other ones." Leading with "you're capable of more" before someone has shown up, communicated, or owned a mistake isn't grace. It's an excuse handed out early. The order is what makes the belief credible when it finally arrives.
The drill below is the paper version of this. It won't fix the concentration by itself. It will make the concentration visible enough that you can't keep not-deciding about it.
Map every recurring decision, approval, account, and relationship in your company, then run the bus testAsking what would break in the business if one key person were suddenly unreachable., asking what breaks if one key person goes unreachable, on each one.
You did it right if at least one item that only you could do last month is now documented or owned by someone else, and you can prove it by not being in the loop when it next runs.
The drill above is the paper on-ramp. The built version does the same inventory and bus testAsking what would break in the business if one key person were suddenly unreachable., then renders it as a hub diagram: people as nodes sized by how much routes through them, with the highest-severity concentrations pulsing so you can't miss them.
Next in this section: Own the Intelligence Layer, what your company keeps when the people and the tools around it change. The habit of capturing judgment you just practiced on your team, the magic-wand question, the touch-point meeting, applies just as directly to what you capture from a model vendor you don't control.