Series: AI × Supply Chain build breakdown
Stack: Claude · Google Docs
The problem
At a small company the process for everything is whoever happens to do it. Doesn’t scale, and it’s fragile: that person’s out sick and the task just stops. SOPs are the obvious fix and the thing nobody ever finds time for.
What I do
Less a build, more a habit. Instead of sitting down to “write an SOP”, I talk Claude through how I actually do the thing, out loud, messy, and it drafts the structure: purpose, steps, the gotchas, what to do when it breaks. I edit for voice and the team-specific bits and drop it in our shared docs. The boring part, turning a brain-dump into a clean doc, is exactly what AI is good at.
Where it failed
First SOPs were too long and too generic. They read like a manual nobody opens, which is worse than no SOP at all. Fixed it with format discipline: short, action first, verbs bolded, the one or two real gotchas called out and nothing else.
And the judgement stays mine. AI gives a great first draft, but deciding what’s actually load-bearing versus filler is the operator’s job. Keep that bit human.
Leverage rating: 3 / 10
Per SOP, small. But standardising how a tiny team works is the quiet thing that lets it punch above its size.
Related: this is the kind of thing I teach in AI for Ops training. Email me for the prompt I use.