Every broken process gets diagnosed the same way. Someone points at the spreadsheet nobody updates, the workflow everyone routes around, the form half the team ignores, and calls it a tooling problem. Add automation. Redesign the interface. Ship a better dashboard. Problem solved.

It's rarely a tooling problem. A spreadsheet doesn't stop getting updated because updating it is hard. It stops getting updated because someone, at some point, decided it wasn't worth their honesty - because nobody read it, or acting on it never changed anything, or the effort stopped being rewarded. That's not a system failing mechanically. That's a person who quietly stopped trusting it.

Most operational fixes skip this question entirely. They treat the system as a piece of infrastructure - logic, inputs, outputs - and ignore that every system is also a record of how much people currently trust it. You can rebuild the interface, automate the reminders, add validation rules, and none of it will matter if the underlying trust never comes back. You'll just get a better-looking version of the same abandoned process.

This matters even more with AI and automation, because the instinct there is even stronger: replace the manual step, remove the person, assume the friction was purely technical. Sometimes it is. But often the "friction" was a person compensating, quietly, for something the system got wrong - catching an error before it became a problem, adding context a form couldn't capture, exercising judgment nobody designed for. Automate that away without understanding why it existed, and you don't remove friction. You remove the thing that was holding the system together.

Before fixing any broken system, the more useful question isn't "what's wrong with the process." It's "who stopped believing in this, and why." That question is slower to answer. It's also the only one that leads to a fix that actually holds.