I wrote today, in Spanish, that it took me more than 30 years to understand that not every table deserves you sitting at it. In a Salesforce org, that idea becomes very practical: your judgment is part of the decision, or you are only there to implement what was already decided.
This is not a guide to quitting. It is a way to stop explaining the same risk for months, with nothing written down, until you get tired of talking to a wall.
The Spanish original is here: https://discover.egafutura.com/no-todas-las-mesas-merecen-que-te-sientes/?utm_source=devto&utm_medium=referral&utm_campaign=amplify
1. Write the risk where the decision lives
If a validation rule, a Flow or a status change is going to break a process, do not leave the warning in a chat thread. Put it on the record: a comment on the work item, the description of the Flow, or the approval task. Two sentences are enough: what can go wrong, on which object, and what you would test before going live.
That does not guarantee anyone will listen. It does mean the conversation cannot later restart as if nobody had flagged it. If the same warning came up in three meetings, it belongs on the object or the work item, not only in the memory of the person who raised it.
2. Record who decided, and what they chose instead
When the recommendation is discarded, write down the decision: who made it, on which date, and which alternative was chosen. In Salesforce that can live in a work-item comment, an approval field, or a short knowledge article tied to the process.
The point is not to be proven right later. The point is that the team can review the judgment, not only the outcome. If that Flow later fails in production, it will matter that the discarded alternative, and the reason it was discarded, were written down.
3. Set a personal limit before you keep pushing
The hard part, and the one that took me the longest to learn, is this: not every table deserves another month of your time. If you have documented the risk, explained it, and the answer stays the same, ask whether that table wants your judgment or only your execution.
Yesterday I published a practical checklist for reviewing an AI-generated Salesforce change by listing what else already runs on the same object. Today's point is the human side of that same craft: knowing when to stop proving your value to people who have no interest in recognizing it.
