Leadership & Transition

The Operating System Walks Out the Door

By John J. BakerJuly 28, 20265 min read

Nearly half of the people who own American businesses are old enough to be thinking about the end of the run. Most of the preparation that follows goes into the deal: the entity, the books, the terms, the tax position. That work matters, and it generally gets done well, because there is an entire profession pointed at it.

The part that goes wrong is somewhere else.

I wrote a framework for reading transition risk that opens on this line: when a founder-led company changes hands, you are not replacing a person, you are replacing an operating system nobody wrote down. That was the diagnosis. This piece is about what has actually changed since, which is that the operating system can now be written down without asking the person who holds it to sit down and write.

The handoff is already underway

Not a demographic forecast. A transfer of ownership in progress right now, across the businesses that employ people.

US businesses owned by someone 55 or older: 2.9MUS businesses owned by someone 55 or older
Employees those businesses support: 32.1MEmployees those businesses support
Of employer businesses have no clear succession plan, or are unsure: 26%Of employer businesses have no clear succession plan, or are unsure

Roughly half of owners surveyed intend to exit within five years.

Sources: Project Equity for the count of businesses owned by people 55 and older and the employees they support; the Gallup Pathways to Wealth Survey (fielded September to October 2024, 1,264 owners) for succession-plan readiness; the Exit Planning Institute for the five-year exit intention.

The layer that does not transfer

A business hands off in three layers, and the fidelity collapses as you go down.

The legal and financial shell moves cleanly. The written process moves partly, and there is a catch inside that worth noticing: the procedures that got documented tend to be the ones that were simple enough to document. Writing something down is easiest when the rule is clean, so the written half of the job is, almost by definition, the easy half.

Then there is the third layer. Why that number is light. Why this client gets a phone call and that one gets an email. Which job to walk away from even though the margin looks fine. That is the layer that actually runs the business, and it is the layer that transfers worst.

What actually survives a handoff

Three layers. The one that transfers worst is the one that runs the place.

1. The legal and financial shell
about 95% survives
Lawyers and bankers. Entity, contracts, balance sheet, the terms of the deal. This is the part everyone prepares for, and it is the part that transfers cleanly.
2. The written process
about 45% survives
Whatever got documented. The SOPs that exist tend to cover the half of the job that was straightforward enough to write down in the first place.
3. The judgment
about 10% survives
One person, undocumented. Why that number is light. Why this client gets a call and that one gets an email. Which job to walk away from. None of it is written anywhere.

An illustration, not a measurement. The proportions are there to show the shape: preparation concentrates on the layer that was never really at risk.

This is not a quirk of small companies. In manufacturing, about a quarter of the workforce is 55 or older, and 97 percent of manufacturers report concern about the brain drain arriving with retirement. The skilled trades are tighter still: roughly 400,000 unfilled positions today, on a path toward two million by 2033. The people who know the most are the people closest to the door.

Why every previous attempt failed

Documenting institutional knowledge has sat on somebody's list at every company for thirty years. It almost never gets finished, and the reasons are structural rather than personal.

The person holding the most knowledge has the least available time. Writing a manual is a large, unpleasant, low-urgency project competing against a full week of urgent work, and urgency wins every time. Worse, most of what you are trying to capture is tacit. Ask a twenty-year estimator for the rule and frequently there is no rule. There is a pattern learned across a thousand jobs that surfaces as a feeling that the number is light. They cannot write it down because they cannot state it.

So the task was never really to write it down. It was to get something out of a person who cannot articulate it and does not have an afternoon.

What changed

Capture no longer requires the expert to be the author.

You record the work while it is happening, the estimator walking a live bid and talking through a real number on a real job. AI turns hours of that into a written standard with the reasoning attached to each call rather than stripped out of it. Then the expert reads the draft and corrects it.

That third step is the whole unlock. Nobody has time to author a manual. Everybody has time to tell you a draft is wrong, and exactly why. Editing is a smaller job than authoring, it fits in the gaps of a real week, and it reaches precisely the tacit knowledge that direct questioning never gets to, because people who cannot state a rule can almost always spot a violation of one.

Capture without asking anyone to write

The expert stops being the author and becomes the editor. That is the part that was never possible before.

01Them
Record the work being done
Not an interview. The estimator walking a real bid, talking through a real number, on a real job. Judgment only shows up in contact with an actual decision.
02AI
Draft the rule from the recording
Turn hours of talking into a written standard, with the reasoning attached to each call rather than stripped out of it.
03Them
The expert corrects the draft
This is the whole unlock. Nobody has time to author a manual. Everybody has time to tell you a draft is wrong, and exactly why.
The unlock
04AI
It becomes the standard
Something a new hire can be trained against, a decision can be checked against, and an automation can be judged against.

Every previous attempt at this failed at step three, because it asked the busiest person in the building to sit down and write. Editing a draft is a different job entirely.

I have written before that an AI loop stops dead at the point where the standard was never written down. This is that same seam approached from the other side. Capturing the standard is not only succession work. It is the prerequisite for nearly everything else you would want to automate, which means the effort pays twice.

What this does not do

The limit is worth stating plainly, because this is an easy thing to oversell.

Capture does not produce a successor. It produces a written standard, and a standard is not judgment. Judgment still has to be rebuilt inside a person, through reps, on real jobs, with someone experienced watching. What changes is the ramp. A new estimator working from a documented standard with the reasoning intact gets there considerably faster than one reverse-engineering it from nothing, but they still have to get there.

It also does not survive being left until the end.

When you run it decides how much you get

This is work you do while someone is still in the seat, not work you do on the way out.

1. While they are still doing the work

The only window where judgment surfaces at all, because it only appears against a live decision.

2. During the notice period

You get the outline of the job. You get very little of the why behind any of it.

3. After they are gone

What is left is whatever happened to be written down, which is the easy half.

An illustration, not a measurement. The green share is how much of the undocumented judgment is still recoverable in each window. It falls fast, and it falls before anyone announces anything.

Judgment surfaces only in contact with a live decision, which means the capture has to happen while the person is still making decisions. Wait for the notice period and you get the outline of the job without the reasoning behind it. Wait until they are gone and what remains is whatever happened to be written down, which is the easy half again.

The version of this that is not about retirement

One more thing, because it widens the piece well past succession.

You do not need to be planning an exit for this to be the most valuable thing you build this year. The same capture removes key-person risk on an ordinary Tuesday. It is what lets you empower a manager to run something without you rather than remaining the escalation point forever. It is what makes a business sellable rather than merely profitable, because a buyer is pricing how much of the operation is portable. And it is the difference between a company that absorbs one person taking a long vacation and one that does not.

An exit just puts a date on a problem the business already has.

Where ClearOak comes in

This is the work we do with owners, and with the operators who back them: finding where the business actually lives in people rather than in systems, capturing that while there is still time to capture it, and building the automations the standard makes possible.

If there is one person in your business whose departure would genuinely hurt, and you cannot point to where their judgment is written down, that is the conversation to have. Schedule a call and bring that person's name.

Prefer email? Reach me at john@clearoakconsulting.com and tell me which seat you would least want to have to refill. That is usually the right place to start.

Back to Frameworks