The formula is short: (hours saved × loaded cost × frequency) minus (build cost + run cost + maintenance + the cost of errors at automated speed). That last term is the one vendor decks omit, and it's the one that decides the project.
Automation doesn't just accelerate work; it accelerates mistakes. A process with a 2% error rate handled by humans becomes a 2% error rate at machine throughput — unless you buy detection and rollback with part of your budget. Sometimes that price flips the equation negative, and the right answer is a checklist, not a pipeline.
We walk away from roughly a third of automation requests after this math. The projects that survive it ship with confidence, because the spreadsheet — not the demo — made the case.