Reading Your Own Workflows
This is a member-only chapter. Log in with your Signal Over Noise membership email to continue.
Log in to readModule 4 · Section 2 of 5
Reading Your Own Workflows
Most people, when they first look carefully at how they work, find a strange mix: some things are genuinely manual because they need to be, some things are manual because they were never set up otherwise, and some things are half-automated in ways that create more friction than they save.
A useful exercise is to pick your ten most time-consuming recurring tasks and map each one to where it currently sits and where it could sit.
Fully manual, done by hand each time. This is fine for tasks that genuinely require your judgement, expertise, or specific relationships. It’s worth questioning for tasks that are mostly mechanical with a small judgement component — those are candidates for delegation where you review the output rather than produce it.
Manual but with AI assistance. You invoke Claude for each instance but haven’t turned it into a reusable skill. This is the right place to be while you’re still learning what good output looks like. Once you’ve done it enough times to know, turn it into a skill.
Skill-based, invoked when needed. You’ve got a skill that handles the task reliably. You run it when you need it. This is the right permanent state for many tasks — not everything needs to be autonomous.
Automated, runs on a schedule or trigger. Fires without your involvement. Right for tasks that need to happen on a rhythm regardless of what you’re focused on — briefings, health checks, routine processing.
Automated but broken or ignored. Runs but produces output nobody looks at, or produces unreliable output that requires constant manual fixing. This is the danger zone — automation that looks like it’s working but isn’t. Worth finding and either fixing or removing.