I’ve shipped automations that looked great in a demo and stalled in week two. The code wasn’t the problem. Nobody owned the new habit, training was a forward, and the old spreadsheet still felt safer.
Name the owners before go-live
Every new step needs a person, not a department. Who starts the request? Who approves? Who closes exceptions? Put names on the SOP. Ambiguity is where old processes sneak back in.
Train on the real cases
Walk through last month’s messy tickets — not a clean happy path. People remember friction. If the automation handles the awkward cases, trust builds faster than any slide deck.
Kill the parallel path
As long as the old form or shared drive stays open, adoption stays optional. Set a cutover date, archive the legacy route, and keep a short exception channel for genuine blockers.
Measure adoption, not just uptime
Track how many cycles still go offline. If that number isn’t falling, the rollout isn’t done — even if the script runs clean. Automation success is behaviour change with a system behind it.