A structural engineering consultancy has spent eight years earning a particular kind of reputation: when its name is on the drawings, certifiers and builders stop asking questions. The firm employs four junior and mid-level engineers who now do most of the drafting and the load calculations. But every set of structural drawings still crosses one desk before it reaches the certifier. The principal engineer reviews each one personally. Not a sample. Every submission.
When the partners raise the obvious question, that the firm could take on twice the work if someone else were trusted to sign off too, his answer doesn't change: no one else can do it like he does.
He isn't wrong, exactly. More than once he has caught something in a last pass that a careful, competent junior engineer missed entirely. He has a feel for which assumptions are merely conservative and which ones are quietly dangerous. Clients who have worked with the firm for years ask for his name specifically, not the firm's.
But "not wrong, exactly" is an odd place for a business to still be standing after eight years.
The Sentence Gets Dismissed or Believed Whole
Most advice aimed at this situation picks one of two unhelpful positions. The first treats the sentence as ego: the founder doesn't trust the team, won't let go, needs to learn delegation. The second treats it as simply true and stops the conversation there, because arguing with genuine expertise feels presumptuous.
Neither position asks the only useful question. "No one can do it like me" is rarely one claim. It's two, folded together so tightly they sound like a single statement.
The first claim is that the judgment involved is genuinely scarce, built from years of pattern recognition that took real time to develop and can't be handed over in an afternoon. The second claim, usually smuggled in beside the first without anyone noticing, is that this judgment has simply never been asked to travel beyond the one person who holds it.
Those are different problems. The first deserves respect. The second deserves a design decision.
What the Business Owes a Genuine Expert
This is the harder version of founder dependence, because the usual fix doesn't apply cleanly. In many businesses, constant founder involvement really is just habit, left over from the early days and never corrected. Here, the involvement is often the reason customers choose the firm in the first place, which makes the two situations easy to confuse and dangerous to treat identically.
Customers aren't always buying a defined deliverable. In expert businesses, they are frequently buying judgment applied to a situation that genuinely matters to them, and no checklist fully substitutes for that. This is the trap *The Designed Business* calls "when expertise becomes the ceiling": the very thing limiting growth is also the reason customers came in the first place.
Stripping the expertise out and replacing it with process would damage exactly what customers are paying for. The useful move is narrower: work out exactly which parts of that judgment are doing something only one person can do, and which parts have simply never been described clearly enough for anyone else to learn them.
The Describability Test
Here is a direct way to tell the two apart, because they feel identical from the inside and need a sharper test than intuition.
Ask the person making the claim to state, out loud, the three or four specific things they're actually checking for when they make the call. Not the general area of concern. The precise signals.
If they can answer quickly and concretely, what looked like an irreplaceable gift is usually an undescribed method. Nobody taught it because nobody ever wrote it down, not because it resists being written down. That's a design gap, not a talent gap, and a design gap transfers.
If the honest answer is closer to "I just know when something's off," take that seriously rather than dismissing it as a cop-out. Some judgment really does resist full description. It's built from thousands of small comparisons the person has made without consciously cataloguing them, and asking them to produce a checklist on demand will usually get you a worse version of the real thing. That part is a genuine scarcity, and trying to clone it badly can be more dangerous than simply relying on it.
The test doesn't produce one answer for a whole person. It produces a different answer for each specific judgment call, which is exactly why "just delegate it" and "he needs to let go" both miss what's actually happening. Most situations contain some of each.
Running the Test Against the Calculations
Pressed on what he's actually checking for when he reviews a set of drawings, the principal engineer names four things without much hesitation.
He checks whether the soil-bearing assumption used in the foundation design matches the geotechnical report, rather than a safer default value pulled from a template. He checks whether every critical connection has a genuine second load path, not a theoretical one that only works if every other element performs exactly as drawn. He checks whether any assumption buried in the calculations would quietly become false if site conditions varied even slightly from what was surveyed.
Those three are describable. None had ever been written down as a checklist, because nobody had ever asked him to separate "what I personally check" from "what I generally supervise." Once named, a second senior engineer can be trained to run exactly the same three checks on the majority of straightforward projects, with the principal's review reserved for anything that trips one of them.
The fourth thing he names is different: a sense, on the harder or more unusual projects, that something about the whole design doesn't sit right, even when every individual number checks out. He traces it, when pushed, to a handful of near-misses from years earlier that he's never consciously catalogued but clearly still carries. That part doesn't reduce to a checklist. It's the genuine scarcity inside the claim.
The firm had been choosing between two extremes: him reviewing everything, or eventually reviewing nothing. Neither was ever the real choice on offer. The describable checks move to a second qualified reviewer, freeing most of his time. The real pattern recognition stays exactly where it already works, reserved now for the smaller set of projects actually complex enough to need it.
What Changes Once the Two Are Separated
Nothing about this reduces the principal to a procedure, and nothing about it extends his unreviewed say over everything forever. Those were the only two options the firm had been choosing between, and both were wrong, because they assumed the whole claim had to be entirely true or entirely false.
The useful move is smaller than either option: name what's describable, write it down, and hand it to someone qualified to run it. Let the genuinely singular part stay personal, on purpose, instead of by accident, reserved for exactly the situations where it earns its place.
This is also where a business decides what it's actually selling. A firm that correctly identifies an irreducible core of expertise can design review structures, minimum information requirements, and escalation triggers around that core, protecting it rather than pretending it doesn't exist. That's a different business from one that simply hasn't looked closely enough to know whether its bottleneck is real or assumed.
A Few Questions Worth Running Honestly
Pick a judgment call in your business that currently only you make. Can you state, out loud, the three or four specific things you're actually checking for, or does the honest answer stop at "I just know"?
If you can describe it, what has stopped you from writing it down and testing whether someone else can learn it?
If you genuinely can't describe it, what structure, a review gate, a minimum information standard, a defined escalation trigger, could protect that judgment's place in the business without requiring you to clone it?
The Danger Isn't the Sentence. It's Never Testing It.
"No one can do it like me" will sometimes be entirely true, and a business that hears it and immediately assumes ego is as wrong as a business that hears it and never asks another question. The sentence itself isn't the warning.
The warning is a business that has never actually tested which half of the sentence is carrying the weight. Some of what looks irreplaceable is, and deserves protecting rather than arguing with. Most of what sounds irreplaceable, in practice, turns out to be something nobody had gotten around to describing out loud.
The difference between those two businesses isn't talent. It's whether anyone ever asked the question precisely enough to find out.
Before you decide whether your own version of "no one can do it like me" is true or just untested, run the describability test against it honestly. Pick the one judgment call in your business that currently only you make, and try to say, out loud, the three or four specific things you're actually checking for. Whatever you find out is worth more than another debate about delegation.