This one is the full method, not an overview. Copy it, build it yourself, keep it. The hard part was never the idea.

Most of what a small business knows lives in exactly one place: somebody's head. Why you fired that supplier. Which customer needs the invoice split. What you tried in 2024 that did not work. The reason the deposit is forty percent and not thirty.

None of that is written anywhere, and all of it is load-bearing. It leaves when the person leaves, and it goes quiet even while they are still here, because nobody can recall on demand what they decided fourteen months ago.

Below is the whole method. Not a summary. If you follow it you will have working organizational memory, and you will not need me for it.

Why the obvious approaches fail

A wiki fails because nobody knows which page is current. A shared drive fails because it is a pile with no distinction between what is true now and what was true once. A chat history fails because it is chronological and unsearchable in the way that matters: you can find when something was said but not what was decided. A notebook fails when the notebook leaves the building.

Every one of these fails for the same underlying reason. They mix current state with history, so reading them requires already knowing which parts are still true. That is the exact knowledge you were trying to store.

Current state and archive must be physically separate.This is the whole trick. Everything else is bookkeeping. If a reader has to judge whether a file is still accurate, the system has already failed.

The structure

Two folders. That is genuinely it.

01

current/

Everything true right now. Small files, one topic each. If a fact in here is wrong, it is a bug and gets fixed immediately. A new person can read this folder top to bottom and be caught up.

02

archive/

Everything that was true once. Nothing gets deleted, because the reasoning behind a superseded decision is often the most valuable thing you own. Files move here when they stop being current; they never get edited after the move.

When something changes, you do not edit history. You update the file in current/ and move the old version to archive/ with a date on it. The current folder is always small. The archive grows forever and that is fine.

What goes in a file

Four things, every time:

01

The date

When this became true. Not when the file was created. Undated organizational memory is barely better than none, because you cannot tell what supersedes what.

02

The source

Where it came from. A conversation with a named person, an email, an invoice, a decision made in a meeting. A fact with no source cannot be checked when it is later disputed, and it will be disputed.

03

The decision or fact itself

Written so someone who was not there can act on it. Not shorthand. Not "the thing we discussed."

04

The reasoning

Why. This is the part everyone skips and the part that pays. A decision without its reason gets reversed by the next person who does not know what it cost you to learn.

The rule that keeps it alive

Write the file when the decision is made, not later. Later never arrives. This is a thirty-second habit at the moment of deciding, and it is the only discipline the system requires. Every other approach dies here, at the intention to write it up properly at the end of the week.

If you only take one thing from this article: at the moment you decide something that will still matter in six months, open a file, date it, write the decision and the reason, save it. That single habit is worth more than any tool.

Where AI fits, and where it does not

It helps in exactly two places. First, drafting: paste in the messy thread and get back a structured file with the four fields, which you correct. Second, retrieval: ask a question in plain language across the folder rather than remembering filenames.

It does not decide what is true. It does not resolve contradictions between two files. It does not maintain the folder for you. If nobody writes files, there is nothing to retrieve, and no amount of clever tooling fixes that.

Why give this one away

Because the method is not the hard part. The hard part is the thirty-second habit at the moment of deciding, and no tool sells you that.

If you read this and build it, good. You should. If you read it and recognize you will not maintain it without something enforcing the shape, that is a fair self-assessment and worth a conversation. Either way you now know exactly what the thing is, which is more than most vendors will tell you.

Follow the connected workflow