How to Document What You Know So It Doesn't Leave With You
Start with one decision you make on instinct, narrate it out loud to Claude, and let it interrogate you until the instinct becomes a written rule someone else can follow. Do that for two hours a month and in a year you'll have 24 documents that capture the judgment no manual ever holds. That's the whole method. The rest is showing up for the second hour.
When a senior person walks out the door, the playbooks stay. The process manuals stay. What leaves is the other thing: the pattern recognition built over 30 years, the sense of when to break your own rules, the read on a counterparty you get in the first ten minutes of a meeting. None of it is written down anywhere, because it never felt formal enough to write. It just felt like knowing.
If you're inside ten years of a real transition (a retirement, a sale, a handoff, a pivot into advising), capturing that knowing is one of the highest-return things on your desk. And for the first time, it's not a slog. Claude won't supply the expertise. But it's very good at the one job that used to make this impossible: asking the questions you'd never think to ask yourself.
Why tacit knowledge is so hard to transfer
The most valuable part of it was never explicit to begin with. Organizational researchers split what you know into two buckets. Explicit knowledge is the stuff you can write in a manual: criteria, specs, formulas. Tacit knowledge is the stuff you only know through having done it ten thousand times. Explicit transfers fine. Tacit usually doesn't transfer at all through standard documentation, which is exactly why succession plans capture the easy 20% and miss the part that mattered.
Take a commercial real estate underwriter who's been pricing risk for three decades. Her explicit knowledge, the underwriting criteria, fits in a manual any junior analyst can read. Her tacit knowledge doesn't: the way a developer's language tightens when they're overextended, the submarket signals the models don't catch, the gut call on a sponsor she forms before the deck is open. That's worth more than the manual. And it lives in exactly one head.
Documenting what you know is really one task: dragging that intuition into a form someone else can actually use. It feels obvious from the inside, which is the trap. "I just read the room" is not a transferable asset. The reasoning underneath it is.
How AI pulls the knowing out of you
The method is plainer than it sounds. You sit down with Claude open and start talking. A simple opener is: "Claude, I'm going to describe a recent [deal/case/launch]. Your job is to keep asking questions until you can write a one-page decision guide someone else could follow." What happens when a deal goes well? What are you noticing that nobody else in the room is? When did you call something wrong, and what did that wrong call feel like before it went sideways? What do you ask in the first five minutes of an engagement that tells you most of what you need to know?
On your own, you hit a wall fast. The knowledge feels too obvious to be worth saying. That's where Claude is useful. It doesn't bring the expertise; it just refuses to accept "I just know" as an answer. Give it your narration and it pushes back: Walk me through the last three meetings where the room read differently than you expected. What specifically tipped you off? How did you recalibrate? That kind of follow-up is what converts a vague instinct into a stated criterion.
You answer honestly. Claude asks the next question. You wouldn't produce this alone, and Claude can't produce it without you.
The four things actually worth documenting
Most people try to document everything and finish nothing. Don't. There are four domains where capture pays off, and the rest is noise.
| Domain | What it captures | Why it's usually lost |
|---|---|---|
| Decision frameworks | The mental models you run on ambiguous calls, the ones you'd call "instinct" | You never wrote them down because you don't think of them as frameworks |
| Relationship maps | Who knows what, who has authority vs. influence, where decisions really get made | It's the first thing an org notices it's missing the week you leave |
| Failure archives | What went wrong, why, and the lesson. Your most useful knowledge by far | Documentation captures best practice; almost nobody captures the scar tissue |
| Judgment criteria | The factors you weight heaviest: when you push back, when you escalate, the threshold | It's the essence of senior expertise and feels too obvious to spell out |
A general manager who's run seven product launches has a framework for when to hold a launch and when to push through, even if she only ever called it instinct. A project lead who's delivered 40 implementations carries more value in what went sideways than in the standard methodology. The failure archive is the one almost everyone skips, and it's the one I'd fight to keep.
The workflow: two hours a month, not a heroic project
I used to think this had to be one big push: block off a week, dictate everything, hand off a binder. That was a mistake, and I've watched it fail in practice. People burn out around hour six and the binder gets skimmed once. Treat it as a recurring practice instead.
Block two hours a month on your calendar. Call it the "Two-Hour Transfer." Each session, pick one domain, project, or decision type. Narrate it to Claude: what happened, what you saw, how you decided, what you'd do differently. In my experience, that's 90 minutes of narration and 30 minutes of editing what Claude structures back. Then ask it to structure the narration into a working document: observations, decision criteria, lessons, and the questions worth asking the next time a situation like this shows up. Twenty-four sessions, 24 documents, each one a slice of expertise that otherwise lived only in your head.
Here's a typical pattern I see with attorneys, anonymized but true to the shape of the work. An M&A lawyer used this to document her pre-LOI diligence checklist with the reasoning behind every item, the red flags she watches for in target management, the three deal structures she defaults to and when each fits, and the negotiations where walking away was the right call, including how she knew. That corpus, reasoning included, ended up as five short memos her associates now use before every deal. It's her judgment, made portable.
Is it too late to start this in your 50s?
No. The sequencing changes, not the value. If you're inside two years, narrow the scope: document the three or four decision types that would cost the most to lose, and skip the comprehensive ambition. A focused capture done in four months beats a perfect one you never start.
And this matters more at 50+ than it did at 35. Your expertise is the moat: AI can summarize policies, but it can't replay the 400 deals, audits, or projects you've actually lived through. Documenting it isn't admitting you're on the way out. It's making the moat visible so you can charge for access to it. The professionals who negotiate the best exits are usually the ones who can say "here's a documented system for how this organization makes decisions, and here's how I'll train your next generation to run it." That's a far stronger position than "you can keep buying my time."
Why most senior people resist this (and why they're wrong)
I hear the same three objections over and over. The first is the quiet imposter voice: is what I know really that valuable? Almost always yes. It just never feels that way from the inside, because the hard-won stuff has become automatic.
The second is the fear that writing it down makes you replaceable. This one's backwards. The professionals whose expertise stays invisible are the easiest to walk past at renewal time. The ones who've made their judgment legible and systematic are the ones who land the board seats, the advisory retainers, the teaching engagements, because they can show what they know instead of just asserting it.
The third is real: it takes time and focus you don't have. Which is the whole reason the AI piece changes the math. The interrogation, the structuring, the synthesis: that's the labor that used to demand a week on a mountain with a recorder. Now it fits in a standing two-hour block on your calendar.
Frequently asked questions
How detailed should the documentation be?
Detailed enough that a smart person who is unfamiliar with the context could execute without calling you. When you write "it depends," stop. Add what it depends on, that's where the value is. Three examples beat ten principles.
Should I share this with my current employer or keep it?
Depends on your goal. Some people document for internal succession, some to build intellectual products they own, plenty do both, with real care about what belongs to the organization versus what's your own hard-earned judgment.
How much can AI do without me?
Almost none of the substance. The expertise is yours, full stop. What Claude handles is the questioning and the synthesis: asking the next sharp question and turning your answers into a structured document. The combination produces something neither of you makes alone.
What format works best?
Fit the format to the use. Internal succession docs work as playbooks. Teaching material works as frameworks with examples. Consulting or advisory material works as case-based analyses.
Can you over-document?
Easily. The fix is to constrain the format. A three-page decision framework with real examples beats a 40-page manual nobody opens. Aim for useful, not exhaustive.
What to do this week
Open a chat, name one decision you make on instinct, and tell Claude to keep asking until you've stated the rule out loud. One session. If it produces a single page you could hand to a successor without needing to walk them through it, you've proven the method. Do it once a month and by year-end you'll have 12 documents nobody else could write.