You started the company because you were good at the work. So you did the work. All of it. Sales, delivery, invoicing, the thing that broke at 9pm on a Tuesday. Doing everything wasn’t a failure. When you’re one person with an idea, doing everything is the job.
Michael Gerber gave the next part a name. In The E-Myth Revisited he calls it the technician’s trap: you’re so good at the work that you keep doing the work, and you never build the business that does the work without you. That’s the quiet problem with “doing everything was the job.” It stays the job while the company outgrows it. You’re now running a 12-person business the way you ran a 1-person one, and the whole thing still moves at the speed of you.
Then it gets heavy enough that you reach for the one lever everybody reaches for.
Hire someone.
But there’s a second way to get work off your desk. Picture the one task you keep dropping already handled, running without you, with no new salary attached. Not a hire you spent three months making. A seat that started as a single job you proved in about a week. That’s the other lever, and most founders never reach for it.
Why every bottleneck in a small company routes back to you
Watch where the work actually stalls in your company for a week. Trace it. Almost every stall ends at the same desk. Yours.
The proposal that’s been sitting for six days because you haven’t written it. The hire who can’t move until you sign off. The client thing that only you know how to answer. That’s not a discipline problem and it’s not a people problem. It’s a design problem. You built a company where the important decisions and the routine ones both run through one person, and that person is out of hours.
The reflex is to add a person. More hands, more throughput, less on your plate. It’s not a crazy instinct. It’s just the only one you were ever handed. And at 5 to 50 people it’s the slow, expensive way to fix a problem you could solve a different way.
From operator to owner: what changes as your company grows
Here’s the shift almost nobody makes on purpose. Your job is supposed to change as the company grows. Early on, your job is to do the work. Later, your job is to build the thing that does the work. Gerber’s version: work on the business, not in it. Operator to owner.
Most founders never make that shift, and it’s not because they’re lazy or scared. It’s because they only own one tool for making it. Delegation, and delegation has always meant another salary.
Delegation is the growth lever, not a nice-to-have. Gallup looked at 143 Inc. 500 CEOs and found the ones with high delegator talent posted a three-year growth rate 112 percentage points higher than the ones without it. The founders who let go grow faster. The ones who hold everything cap out at their own calendar.
So the instinct is right. Get the work off your desk. The problem is the price of the only lever you’ve got.
The real cost of a hire is the bottleneck you chose not to build out of
Let’s be honest about what a hire actually costs, and let’s be fair about it too. This isn’t hire-versus-do-nothing. You’ve already decided this work needs to get done by someone who isn’t you. The real question is how you get it done. So put the two options next to each other.
Making the hire you were about to make: It takes about twelve weeks to fill (83 days) and costs 30% to 200% of first-year salary once you count recruiting, ramp, and the risk it doesn’t work out (Management.org; US Department of Labor; SHRM). And the part nobody prices in: for the first few months the work still routes through you, because you’re the one training them. If it’s the wrong hire, you’re back at week zero, plus severance.
Building a seat instead: One bounded task to start. Not a role, not a department. A single job with a clear input, a clear output, and a fast way to check if it’s right. You prove that first job in about a week. If it holds up, you stack the next job on top, and over time those jobs add up to a seat. If it doesn’t hold up, you’ve lost a few days, not a quarter and a salary.
Same work off your plate. One path takes twelve weeks and a permanent line on payroll. The other starts proving itself in a week, before you commit to anything permanent. The real cost of the hire was never just the salary. It was the bottleneck you paid twelve weeks and a headcount to keep, instead of building your way out of it.
How to build your first AI seat: job, then seat, then org
The framework is three words. Job, seat, org.
A job is one bounded task. Sort the inbox and draft the routine replies. Chase the follow-ups you keep dropping. Turn every sales call into notes and next steps. One thing, checkable in a glance.
A seat is a stack of jobs that adds up to the hire you didn’t make. Email triage, plus meeting notes, plus first-draft follow-ups, plus lead tagging, and suddenly that’s not four tasks. That’s most of an operations coordinator, running around you.
An org is every seat you’ve built, working together. You don’t design it up front. You get there by proving one job, then the next, until the seats start adding up on their own.
This is exactly where founders stall, and there’s a number for it. AWS and Techaisle found 34% of small and mid-sized businesses are piloting generative AI, and only 3% have actually integrated it (AWS/Techaisle). That gap is the whole story. Founders get stuck trying to design the AI strategy before they’ve proven a single task. One bounded task working in a week beats a six-week rollout plan that dies in a meeting.
One honest thing, because I won’t sell you a fantasy. A seat is a real build. “Start with one task” doesn’t mean AI is easy or that you should go teach yourself to wire up a pipeline between your CRM and a model at midnight. It means you start narrow so you can prove value fast, on real work, before you bet the company on it. The building is a technical lift, usually for a partner who does this. Your job is to name the task that drains you most and keep your hand on what matters. You shouldn’t be absorbing the technical weight of this alone, and you don’t have to.
What an AI team actually looks like inside a 5-person company
I’ll show you mine, because I built it on myself first.
The first seat I built was follow-through. I’d say yes to something on a call and then the thread would go cold. Not from bad intent, it’s just that closing loops has never been where my energy lives. So I built a teammate named Tess whose only job is chasing those loops down. One gap, one seat. A promise I make on Monday doesn’t quietly die three weeks later anymore.
She wasn’t the last one. There’s a teammate who owns the day-to-day ops grind. One who owns building and shipping software, so my half-finished builds actually cross the line. One who owns getting the work in front of people and keeping the voice consistent. One who translates a rough idea into a real plan. Each one exists because a specific stage of work was draining me, and each has a defined job, not a vague “help with everything.”
In my own build, I show up at exactly five points. Five. I frame the idea, I react to the design wireframe, I make the first go/no-go call, then the second, and I sign off on the ship. Everything between those points runs without me. For a founder who’s used to being in every decision, that number sounds like either a dream or like losing control. It was both, for about a week. Then it was just how the company runs.
Now here’s the part that matters, and it’s not about me.
Jiaona Zhang, the CPO at Laurel and formerly at Airbnb and Dropbox, runs her company on the same cut. She maps every function’s work and color-codes it. Green is the work worth doing more of, the work that made you good, and you protect it. The tedious rest gets handed to an agent. She also found that adding people isn’t free the way founders assume, because every head you add adds coordination cost. This isn’t about cutting your team. Your people are the point. It’s about not adding a head every time the work grows, so your lean crew puts out the work of a company three times its size.
I didn’t automate everything. I kept five seats for myself on purpose. The work I’m actually rare at stays mine: being in the room with a buyer, framing the problem nobody else in the room can see, the judgment call on which idea is worth backing. That’s the point of the whole exercise. You’re not trying to hand off the work that only you can do. You’re trying to hand off everything else so you have room to do it.
So here’s your version. Where does work reliably stall in your company? What’s the one task that drains you every single day? That’s your first job. That’s the seat to build first.
The seat I haven’t finished yet (and why that’s the point)
I’ll tell you the one I haven’t cracked, because a founder telling you he’s got it all figured out is a founder you shouldn’t trust.
My marketing and sales seat isn’t finished. The part of getting KyberFive in front of the right founders that still needs me on camera, in the room, being the face of it, that’s still mostly me. It’s my bottleneck. It’s the exact thing I help other founders build out of, and I haven’t fully built out of it myself yet.
That’s not an accident I’m hiding. My own bottleneck and my next product are the same problem. When I finish that seat, it becomes the next one I can install for someone else, built in public with the receipts to show.
I’m telling you this so you know the honest shape of it. Some seats are built and running today. Some are still being figured out. A founder who claims the whole thing is solved is selling you a finished machine that doesn’t exist. The real version has a gap in it, and the gap is where the next build comes from.
Two years from now
Picture two founders, same revenue today.
One keeps reaching for the only lever he was handed. Every time the work grows, he adds a subscription or a salary. His payroll climbs with his revenue, and he’s still the bottleneck in every important decision, because he never changed how the company works.
The other builds seats. One job at a time, proven in a week, stacked into the hire he didn’t make. Two years out he’s at the same revenue on a fraction of the payroll the other guy piled on, and he’s not the one who’s lean. He’s the one who’s free.
You don’t get there with a strategy deck. You get there by proving the first job, then the next. Name the one task that drains you most. That’s your first job, and where the first seat starts. If you want to find where the drain actually is in your company and which seat to build first, let’s start a conversation.
Frequently Asked Questions
What does it mean to “work yourself out of the work” as a founder? It means your job changes as the company grows. Early on your job is to do the work. Later it’s to build the thing that does the work without you, so the company stops moving at the speed of your calendar. Michael Gerber calls it working on the business instead of in it.
Why is every bottleneck in a small company the founder? Because most founders build a company where the important decisions and the routine ones both run through one person. Trace where work stalls for a week and it ends at your desk. That’s a design problem, not a discipline problem, and adding hours won’t fix it.
Why isn’t hiring the right lever for a 5 to 50 person company? Hiring is slow and expensive, and it isn’t free compared to doing nothing, it’s compared to a hire you were already going to make. A role takes about twelve weeks to fill and costs 30% to 200% of salary once you count ramp and the risk it doesn’t work out. And the work still routes through you until they’re up to speed.
What is the job, seat, org framework? A job is one bounded task with a clear input and output. A seat is a stack of jobs that adds up to the hire you didn’t make. An org is every seat working together around you. You don’t design it up front. You prove one job, then the next, until the seats add up.
What’s the first AI seat a founder should build? The one covering the task that drains you most or the stage where work reliably stalls. In my own company the first seat was follow-through, a teammate whose only job is chasing down commitments so they don’t die three weeks later. Your version might be the inbox, or lead follow-up, or turning calls into next steps.
Does building AI seats mean replacing my team? No. The point is not to cut people. It’s to stop adding a head every time the work grows, so your lean team puts out the work of a much larger one. Your people are the point. The seats cover the hires you would have made next.
Do I have to build the AI seat myself? No, and you shouldn’t try to hand-wire a pipeline between your systems and a model on your own. Starting narrow is about proving value fast on real work, not about AI being easy. The build is a technical lift, usually for a partner who does this. Your job is to name the task and keep your hand on what matters.