Automation projects stall because decision-makers cannot see the number. Yet a solid estimate needs only three inputs: frequency, duration and error cost.
Take one process, measure it for a week: how often it runs and how long it takes. Scale it to a month and add the cost of fixing a failed run.
If that figure does not reach twice the estimated automation cost within a year, this is not the process to start with.
Measurement has a second benefit: at the end you can prove the result rather than defend the next phase on instinct.