Insights
When your best employee leaves, what walks out the door?
Everything that was never written down — which in most small businesses is nearly everything that makes the business run well. The job duties are replaceable; the ad for the replacement practically writes itself. What walks out is the invisible layer on top of the duties: which customer needs the long explanation, which supplier actually answers on Fridays, what that noise on the second floor means, which "standard" job is never actually standard. None of it is in the file. All of it was the reason she was your best.
You can't stop people leaving. You can control how much of the business leaves with them — and the time to control it is now, while they're still here and nothing is urgent. That last clause is the whole problem, so let's start there.
Why does knowledge never get written down?
Because the cost and the benefit live in different months. Writing down what you know costs time today, from exactly the people whose time is most oversubscribed — and pays off on some future day nobody can schedule, when someone quits, gets hurt, or wins the lottery. Urgent beats important; today's job beats next year's resignation, every single day. So the documentation project stays permanently third on the list, and shops that have been burned once are shocked how fast it slides back to third.
There's a quieter reason too, worth saying out loud: experts can't see their own knowledge. Ask your best tech to "write down how you do the job" and you'll get a page of the obvious steps — because the two hundred micro-judgments layered on top don't feel like knowledge to the person holding them. They feel like common sense. They are not common sense. They are twelve years of pattern-matching wearing common sense's clothes, and the standard "have everyone document their job" initiative fails precisely because it asks the fish to describe water.
What exactly is at risk? Take the inventory.
Five categories, and it's worth walking your own roster through them, name by name. Customer knowledge — histories that live in heads, not the CRM: the quirks, the sensitivities, the "always call before noon" rules. Vendor and workaround knowledge — who really honors warranties, the part number that supersedes the catalog, the trick that makes the old software cooperate. Process exceptions — the unwritten rules layered onto every written one: when the standard price doesn't apply, when the standard job isn't. Diagnostic judgment — the pattern library: what that sound, that smell, that customer phrasing usually means. Relationship capital — the accounts that are loyal to a person, not to the shop; when they go, the account wobbles.
Now the uncomfortable row of the inventory: run yourself through it. In most small shops the largest holder of undocumented knowledge — pricing judgment, the whole customer history, every vendor relationship — is the owner. Whatever you feel reading this about your best employee, your business feels about you, multiplied. Every fix below applies double at your desk.
How do I actually capture it — without a documentation initiative?
The initiatives fail; the habits work. Four that survive contact with a real shop, orderable by effort. Capture at the moment of exception. The knowledge shows itself whenever someone says "actually, for this customer we do it differently." That sentence is gold leaving a vein — a thirty-second voice memo at that moment beats the retreat afternoon that never happens. The discipline is one habit, not a project: when the exception happens, it gets captured. Interview, don't assign. Don't ask experts to write; sit with them and ask the water-describing questions — "walk me through the last time this went sideways," "what would a new guy get wrong on this job," "which customers should I never send him to alone." Record it. An hour per person per quarter outruns most binders in existence. Let the tools you already run remember. Job notes actually entered, calls logged, photos attached — the systems you already pay for are half a memory if anyone feeds them. The feeding is the habit that fails, which is exactly the kind of failing a system can absorb. Make the transcript an asset. This is where the current generation of AI earns its keep, unglamorously: recorded walkthroughs and interviews used to be write-only — nobody was ever going to type them up. Now transcription and drafting turn an hour of your tech talking into a searchable, editable draft procedure the same day. The expensive part — extracting what's in the head — was always the labor; that part just got cheap. What it produces should live somewhere you own, in formats that outlast any one tool: this is the SOP-and-knowledge pattern among the six builds, and it's the one whose value compounds quietly for years.
What's this worth in dollars?
Run it like any other leak, with your numbers: what did the last key departure actually cost — the hiring gap, the training months, the mistakes the replacement made that the veteran wouldn't have, the accounts that wobbled? Shops that have lived one don't need convincing; shops that haven't are borrowing against it. The premium on the insurance is a few habits and some hours. The claim it pays out is continuity — and it's one of the few investments here that pays even if nobody ever quits, because a business that can explain itself trains faster, delegates cleaner, and is worth more to anyone who ever looks at buying it.
The judgment call is where to start: whose head, which category, captured how, in what order — sized against everything else leaking in your shop. That ranking is the call that pays, and it's the work I do. Bring me the name you thought of three paragraphs ago: book a conversation.