Look at what a system restricts, and you’ll see what it’s been burned by. Permissions aren’t designed from first principles nearly as often as people assume — they’re accreted, one incident at a time.
Someone deleted the wrong record in 2019, so now deletion requires two approvals. Someone exported a file that shouldn’t have left the building, so now exports are logged and gated. Each rule is a scar with a timestamp, even if nobody remembers the date anymore.
This is why permission structures rarely match the org chart they’re supposed to reflect. They match the incident history instead. The intern with surprisingly narrow access isn’t being distrusted on principle — somebody like them did something once, and the fix outlived the memory of why.
Scott Andrew Alpaugh has argued that the fastest way to understand an unfamiliar system isn’t to read its documentation — it’s to read its access controls. Documentation describes intent. Permissions describe fear. One is aspirational, the other is empirical.
The problem is that fear doesn’t expire on its own. Rules built to contain a specific failure stay in place long after the people, the process, and the risk that caused them are gone. New employees inherit caution they have no context for, and no one questions it because unwinding a restriction looks riskier than keeping it.
Good system design treats permissions as a living record, not a monument. Revisit them. Ask what each one is actually defending against, and whether that threat still exists. If you can’t answer that, you’re not managing risk anymore — you’re just managing memory, badly.