Design the system
The judgment your client’s people carry gets specified as a system — what gets decided, on what basis, and what done right means. Criteria, not a checklist.
For Fractional CFOs
You put a number on it before anyone else calls it a problem. This one is about the client where the number was right and none of your levers reached it.
Recognition
What to listen for
You hear it in the monthly review, usually as the reason a good month didn’t land where it should have. A few things worth listening for:
- “We got the rate increase through. Margin held two quarters, then went right back.”
- “We hired the two people you modeled. The backlog didn’t move.”
- “Utilization looks fine. The work still sits until she gets to it.”
- “We documented all of it last year. They still walk down the hall and ask her.”
- “A buyer is going to ask who runs this without him. We don’t have an answer that survives diligence.”
You can already put a number on every one of those. That is not the same as having something that moves them — and the last one comes with a date.
What I do about it
If you’re hearing those
There’s something I can do about that, and it isn’t a tool your client bolts on.
The reason that number won’t move is specific to how their business actually runs — who decides what, which exceptions get routed where, what “done right” means when nobody is checking. None of it is written down anywhere, which is exactly why the work can’t move to anyone else.
That part doesn’t come off a shelf. Point generic AI at it and you get something plausible that doesn’t fit, and the mismatch surfaces later, inside the work, after everyone has stopped watching for it.
So it’s a build, designed around that one system in that one business.
See one working
Try it on a client you have in mind
That isn’t a claim you should take on faith, so here’s one running.
Tarvi handles the day-to-day operational work inside Admentus — my own business, not a demo. It is the small version of what I just described, grounded in how this business actually works, so when you describe a client it is reasoning against something rather than improvising.
One distinction, because it’s what most people get wrong. This isn’t a chat window onto a general-purpose model — you can open one of those anywhere, describe a business it knows nothing about, and get back something articulate and generic. And Tarvi is one agent. What gets built for a client is several, each with a defined role inside a process that was designed first, with people checking where checking matters.
This one is for you, not your client. Describe the situation you’re looking at, in your own words. There’s no form and nothing to sign up for.
The order matters more than the tooling does, though, and it isn’t obvious. Here’s the clearest example I know.
Detroit, 1906
The math was right
Henry Ford never had an advisor like you. That’s the only part of this I’m making up. Everything after this sentence happened.
It’s 1906, and your client has his thesis worked out. Build a car for ordinary people instead of the wealthy, and get the price low enough that they can actually buy one — volume is where the margin has to come from, not price. It isn’t a vague plan, and he didn’t arrive at it comfortably. He fought his own financial backer over it, and the backer sold out of the company rather than go along.
The thesis is right, and this year it’s working. The Model N sells for around $600 and outsells every other car in America.
Then you sit down for the monthly review, and he puts something on the table the model never accounted for. Orders are coming in faster than the shop can fill them.
The month it stops working
You both reach for what you know
Before you decide anything you go look at the shop, because that’s the job. Every car in it is hand-fitted. A fitter takes a part, files it down, and works it until it seats, because the parts arrive close but never identical and every man has his own way of making them meet. The shop turns out a few dozen cars a day, and only on the days those particular men are on the floor.
So you do what your discipline does, and it’s the right call almost every time you make it. Cost out the unit. Model what a second shift does to it. Tell him the demand supports a higher price, because it plainly does. Ford does what an owner does, and hires more men.
Watch what the hiring actually buys him. Every craftsman he takes on has to be trained by the men already there, so the know-how stays exactly where it has always been, in their hands and their judgment. He’s bought more of the same ceiling at a higher payroll, and the cost per car is exactly where it was.
You’re the only person in that building who can see the whole thing at once. The men on the floor see their own bench. Ford sees demand he can’t fill. You see unit economics that work, an owner doing everything the model says to do, and a cost per car that won’t come down, and you’re the one who says out loud that the problem isn’t the price and isn’t the people. It’s that every car in that shop is still being made by hand, one at a time, by men who each know how in their own way.
That read is correct. It’s also the point where the work leaves your category.
The right order
The car had to change first
What changed was the car.
The Model T arrived in 1908, built from standardized interchangeable parts and designed from the start to be manufactured instead of fitted by feel. A part went in because it was made to go in, not because a skilled man made it go in.
Only then did the moving line work, and it didn’t arrive until 1913, five years after the car it was built for. The method came second because it had to. Run a moving line on hand-fitted parts and all you get is chaos moving faster.
The men who knew how a car went together didn’t lose their jobs in that switch. What they knew moved out of their hands and into the fixtures, the gauges and the dies, and they moved with it. They built the machines, set the tolerances, and handled everything the line couldn’t.
Nobody could have drawn that up in the room in 1906. Not Ford, and not you. It took him five years and a different car to get there.
Your client’s people own it
The two who held that knowledge stop being the work and start owning the system that runs it. The expertise stays; the dependency doesn’t.
Then it scales
AI agents run the repeatable work inside the defined system, so volume stops being a staffing question and margin stops tracking headcount.
Design first, method second. Most firms try it in the other order.
Where I come in
You already did the hard part
That’s where Ford stops being useful to you. He had five years. Your client has this year’s forecast and a number that’s been missing it since spring.
You already did the hard part. You gave them books they can trust, a forecast that holds up under questioning, and an honest read on which work actually makes money. Most firms never get that far, and the ones that do got there because somebody like you made them.
You’ll know the client I mean. The pricing is right, the people are good, nobody is dodging anything, and every line on the report is where it should be except the one that matters.
What’s left sits underneath the number. The work behind it routes through two people’s heads, and there’s no version of it that happens without them. Documenting it is what makes the risk visible in the first place. It doesn’t move the work to anyone else, because a written procedure holds the steps and not the judgment the steps are hiding.
You can see that from where you sit. Neither of you can fix it in the room, though, because the fix isn’t a decision. It’s a build.
The method
What I actually do
I take one business system, the one the constraint actually runs through, and map how it works today under real load. Then I design it so it no longer depends on what any single person carries around in their head, and I build the AI-operated system that runs the repeatable part of it. The people who used to do that work own the system instead of being it.
This fits firms roughly $3M to $15M in revenue, profitable, with a real system already running and one clear place it’s constrained. One system per engagement, not the whole company at once.
It’s the wrong call for a business still finding its footing, for a team that can’t agree on who decides, or for an owner who has already picked the tool they want built. And if what we find turns out to be tactical rather than architectural, I say so, and that’s where it ends.
See the full methodFor you
What referring gets you
There’s no fee and no kickback here. Nothing changes hands, so there’s nothing about the introduction that could look wrong to your client later if it ever came up.
What you get is an answer for a pattern you can already price but can’t build your way out of. Telling a client honestly that the next move is architectural, and knowing who does that work, is worth more to the relationship than guessing at it.
I’m glad to go the other direction too. If a session or a written piece for your group would be useful, I’ll do it. And if you just want somebody to sanity-check a client situation before you decide it’s even worth an introduction, call me for that.
Next steps
When you’re ready to introduce us
However you make the introduction, it goes the same way from there. I follow up once, I don’t chase, and if it isn’t a fit I say so early rather than late. The diagnostic is the entire ask: two hours, $2,500, credited in full toward the engagement if they move forward within 14 days.
Send them the short version
One page written for the owner. Reading it commits them to nothing.
See the one-pager