Business practice
SOP and documentation.
The knowledge your firm runs on, out of two people's heads and into procedures your team actually follows.
Your firm already has procedures. They live in habit, in memory, and in the two people everyone asks when something unusual comes up.
That arrangement works right up until it does not. A resignation, a leave, a busy season, and suddenly the answer to how a critical piece of work gets done is unavailable. What lives only in someone's head cannot be delegated, priced, handed to a new hire, defended in an audit, or given to an AI tool. Every one of those doors stays closed until the knowledge is written down.
The answer is not a binder. Most documentation efforts die because they try to write everything, in language nobody recognizes, and the binder is out of date before it is finished. We write the short list of procedures that actually cost you money when they wobble, in your team's own words, each one with a named owner and a review date so it stays true.
The approach
Three words we work by.
Few.
Not every task deserves a document. We choose what to write by what it costs: the workflows where an absence hurts, an error compounds, or a new hire stalls. A dozen procedures that stay true beat a hundred that decay, and the short list is what makes review possible at all.
True.
Each procedure is written from watching the work happen, in the words of the person who does it, workarounds included. A document that matches reality reads like recognition. A document that describes how work is supposed to happen trains your team to ignore documents.
Alive.
Every document carries a named owner and a review date, because an SOP without both is already becoming fiction. The review takes minutes. What it buys is a library your team can trust, which is the only kind anyone keeps using.
The work
Four workstreams, in sequence.
Dependency audit
A pass through your critical work asking one question of each workflow: if the two people who know this were unreachable for a week, what happens. The workflows where the honest answer is that things quietly stop become the writing list, ranked by what each one costs.
Capture
Each procedure drafted from observation and a short interview, not from a template. Your team reviews and edits the draft so the final document is theirs, in their vocabulary, at the level of detail they would actually consult on a bad day.
Ownership and cadence
Every document gets a named owner, a home your team already looks in, and a review date. One person holds the library as a whole, so the question of whether documentation is current has an answer and an address.
AI-ready structure
The procedures are structured so a new hire and an AI tool can both follow them: explicit inputs, decision points, and definitions of done. Firms that later adopt AI on top of written process get useful output. Firms that adopt it on top of tribal knowledge get confident guesses.
Related reading.
All insights
Operations
The two-person dependency test
Pick any critical workflow and ask one question: if these two specific people were both unreachable for a week, what happens. If the honest answer is that things quietly stop, you have found the real risk register of your firm, and it does not appear in
Method
Why the SOP nobody reads is worse than no SOP
A procedure that does not match reality trains your team to ignore procedures. The fix is not enforcement. It is writing the document from observation of the real work, in the words of the people who do it, so that reading it feels like recognition instead of
Questions
Asked before, answered plainly.
Because we do not document everything. The binder died of ambition: too many documents, written by whoever had spare time, with nobody responsible for keeping them true. We write the short list that actually costs you money, from observation, and every document leaves with an owner and a review date. Fewer, truer, maintained beats complete and abandoned.
Fewer than you fear. In most firms a small fraction of the workflows carry most of the operational risk, and the dependency audit finds them. A first engagement typically produces a working library of ten to twenty documents, not a hundred. The point is not coverage. It is that the ones that matter exist and stay true.
We draft, your team corrects. Watching the work and writing the first version is our job, because asking busy people to start from a blank page is how documentation efforts stall. Your team's job is the part only they can do: reading the draft and telling us where it is wrong. That review is also what makes the final document theirs rather than ours.
It means the procedure states its inputs, decision points, and definition of done explicitly, instead of assuming context a longtime employee carries in their head. That structure costs nothing extra to write and it pays twice: a new hire can follow the document without a translator today, and any AI tool you adopt later has something real to work from. You do not need to be using AI to benefit. You need it the day you start.
Observation costs almost nothing, since watching work is not interrupting work. Per procedure, the ask is a short interview and a review of the draft, usually under an hour of the doer's time in total. The engagement is designed so the people who know the work spend their time correcting, not authoring.