An SOP turns “the way we do this” into something another person, or an agent, can actually follow. It doesn’t make the worker irrelevant. It hands them the expected path instead of a guess.
Think of the recipe card in a kitchen. If it only says “make the sauce good,” it’s useless. A good one names the ingredients, the order, the timing, and the signs that something’s done.
How it shows up
In Benali’s operating model, an SOP often sits underneath a core activity, the named unit of work, and explains how that unit gets done. When the work becomes repeatable enough for AI, the SOP can become part of a skill. The first version usually comes from doing the work manually, noticing the real steps, and writing them down, including the tribal knowledge everyone assumes is obvious. A knowledge base is where SOPs become reusable instead of scattered across chat threads and random docs.
Why you care
A good SOP also marks what judgment stays human: prepare the draft, flag unusual items, wait for approval before sending. The moment repeatable work is written clearly, you can train people faster, delegate to agents better, and stop paying the memory tax every time the work comes back.