do not automate a bad habit.
speed is not always improvement.
if a team sends the wrong report by hand once a week, automation can send the wrong report every hour. if approvals exist only because nobody trusts the process, software can make the distrust move faster without removing its cause.
automation multiplies the behavior it receives.
before building the workflow, watch the work.
why does this step exist? who uses the result? where do people correct the process manually? which exception carries important context? what would happen if the activity stopped for a month?
some tasks should disappear, not accelerate.
a repeated copy and paste may reveal that two systems should share a source of truth. a long approval chain may reveal unclear authority. constant customer reminders may reveal that the original promise was never made understandable.
fix the decision before fixing the motion.
then automate the stable portion. keep unusual cases visible. add a way to pause, inspect, and recover. make failures louder than successes, because silent automation can create damage at a scale no person would tolerate manually.
measure the outcome instead of the volume. fewer clicks mean little if errors rise. faster processing means little if customers spend more time correcting the result.
talk to the people who handled the old process. they usually know which ugly manual step was quietly protecting the customer from a larger mistake.
automation should remove waste while preserving judgment.
the best candidate is a process the team understands, can explain, and would still choose if a machine were not available. software can then make that good practice more reliable.
do not give a bad habit a scheduler and call it transformation.
