At some point, someone on your team leaves or goes on vacation — and you find out the process was them. Not the documentation, not the workflow. Them. The way they remembered to do the thing, the informal checks they ran, the judgment calls nobody else knew were being made. That moment of discovery is useful. It tells you exactly which of your operations are systems and which are people.
The diagnostic question is simple: if this person wasn't available for two weeks, would this process still run correctly? Not 'would someone figure it out eventually.' Would it run correctly, with the right steps, in the right order, with the right evidence? For most recurring processes in growing companies, the honest answer is no — and the reason is almost never laziness or disorganization. It's that nobody designed the process to run without that person.
Documentation doesn't fix this. A Notion page that describes the process doesn't run the process. What breaks in person-dependent operations is execution: no one else knows which step requires a manager sign-off, which attachments are mandatory, which conditions should trigger an escalation. That knowledge lives in the person's head, and when they're gone, it goes with them.
Converting a person-dependent process to a system-dependent one means three things. First, every step has an explicit owner by role, not by name. Second, every decision point with compliance or quality implications has a formal approval gate. Third, every handoff has a defined trigger — not 'send it when you're done' but 'the next task unlocks when this attachment is uploaded.' Those three changes move institutional knowledge from heads to infrastructure.
When it works, onboarding the next hire into a process takes hours instead of weeks. A manager going on parental leave doesn't require a knowledge-transfer marathon. An auditor asking what happened six months ago gets a run export, not a search party. The process doesn't remember anything — it just records everything.