A founder of a 28-person services firm paid a consultant £18,000 three years ago to document the company’s processes. The consultant produced a 230-page operations manual. Nobody reads it. New hires are still asking the same questions the manual was supposed to answer. The founder is still making the same decisions weekly. The manual sits unused on a shared drive.
She is reaching for SOPs again because the COO she hired six months ago needs them, and she wants to know how to do this so the work is not wasted a second time.
Why does the standard write-from-memory approach fail?
The founder tries to articulate processes from memory and produces vague two-page documents that capture the what but not the why. The documentation goes stale within weeks because the work evolves and the documentation does not track the evolution. People revert to asking the founder or improvising. The shared folder becomes a graveyard of abandoned procedures and the firm runs on memory again.
The traditional approach inverts the natural learning sequence. People perform procedures by doing them, learn through repetition and feedback, and articulate the steps last. Documentation written before performance captures abstractions, not the working method. The fix is to record the work as it actually happens, then extract the structure from the recording.
What does the 20 percent look like in practice?
Most owner-led firms have between 50 and 100 distinct procedures. Documenting all of them simultaneously fails on both ends: the documentation team burns out producing material the practitioners do not adopt, and the practitioners cannot read 100 SOPs in any usable timeframe. The Pareto principle holds in operations. Roughly 20 percent of procedures drive 80 percent of operational load and 80 percent of founder time.
In a professional services firm the 20 percent usually contains client onboarding, proposal pricing and approval, vendor selection, staff hiring, performance management, and quality sign-off on deliverables. These are the procedures the founder gets pulled into weekly. They are also the procedures that, if transferred cleanly, free the most founder time.
What happens in inventory weeks 1 to 4?
The founder lives in their calendar for two typical weeks. Either the founder or an assistant records every fifteen-minute block in a spreadsheet, categorised by activity type. After two weeks the data is ranked by frequency and by founder involvement. The output of week one is a ranked list of ten to fifteen procedures scheduled for documentation over the next eight weeks. The remaining 60 to 80 procedures are noted but deferred.
This audit is uncomfortable. Most founders are surprised by how much of their week sits on operational decisions they assumed were already happening elsewhere. The inventory phase usually reveals that 40 to 50 percent of founder time goes on operational firefighting, not on the strategic work the founder thought they were doing. That data is the foundation for everything that follows.
What happens in record weeks 5 to 8?
For each procedure on the priority list, a designated systems champion or subject-matter expert performs the procedure on real customer work while recording their screen. They narrate decision points and explain the reasoning behind them, not just the steps. Tools like Loom, Tango, and Trainual capture the recording and auto-extract a text SOP with screenshots, step numbers, and decision branches. Each recording produces a three to eight-minute video.
The critical discipline is recording on live work, not on a fabricated walkthrough. A walkthrough captures the procedure as it should be; live work captures the procedure as it actually is, including the small judgement calls and exceptions that make up most of the value. The text SOP that emerges is reviewed by the systems champion and the founder, then published to the SOP repository for use.
What is the hand-off test?
The hand-off test is the standard every SOP must pass: a moderately capable new hire, reading the SOP for the first time and with no other guidance, can perform the procedure satisfactorily. If yes, the SOP holds. If they have to ask three questions to get started, the SOP has failed and the founder is still the system.
The test is run literally, not theoretically. A colleague unfamiliar with the procedure attempts it from the SOP alone. Three failure modes appear repeatedly. Documenting at too high a level (prioritise customer requests, without saying how). Documenting steps without decision logic (the how but not the why). Documenting work that is fundamentally judgement (client strategy decisions, creative choices, nuanced relationship calls). Judgement work needs principles or worked examples, not steps.
How does quarterly governance stop the decay?
SOPs decay 1 to 2 percent per month if the underlying work remains relatively stable. By six months a January-written SOP is substantially inaccurate. By the financial year-end it is largely obsolete. The decay reflects a structural feature of operations that change. New software gets introduced. Regulatory requirements shift. Customer preferences evolve. The SOPs do not, unless governance forces them to.
Each SOP gets an owner who is not the founder. The owner conducts a quarterly fifteen-to-twenty-minute review, stepping through the actual procedure against the documented version. Minor deviations are corrected in line. Substantial deviations trigger a re-recording using the same record-and-extract method. The quarterly review is built into the calendar and treated as non-negotiable. Without it, the SOPs sit in the repository unread within twelve months.
Can AI tools compress the timeline?
Recording yourself performing the procedure for fifteen to twenty minutes and feeding it through an AI SOP generator produces a text SOP that is 60 to 80 percent accurate on first pass. Human review is still required, but as refinement, not creation. Practitioner reports from consulting firms using this approach put time per procedure at roughly eight hours total, against forty hours for the manual approach.
The bulk of that saving sits in extracting the structure from the narration. For a firm with eight to ten priority procedures, AI-assisted recording can compress the 90-day sprint to roughly 30 days. The structure stays the same: inventory, record, review and rotate. The compression sits inside the record phase. The hand-off test and quarterly governance still apply. The technology is a multiplier on the discipline, not a substitute for it.
What to do this week
Pick the two procedures the team escalates to you most often. Block fifteen minutes for each and record yourself performing the procedure on the next live request that comes in. Narrate the decisions, not just the steps. At the end of the day you will have two recordings ready for extraction, and the data point of how much time the recording itself takes. That is the seed of the inventory phase.
If the recordings already feel useful before extraction, the 90-day sprint will work for your firm. If they feel like noise, the inventory audit may surface different priorities than the ones you assumed mattered most. Either way the data tells you where to start.
If you would like a second pair of eyes on which procedures to document first, book a conversation.
