Most automation programmes begin with ambition. A platform, a roadmap, a list of systems that “should” be automated. Then the first weeks disappear into setup, and the first useful result is months away.

That’s backwards.

The automations that stick usually start somewhere quieter: a job the team already hates doing. The one that shows up every Monday. The one that only one person knows how to run. The one that breaks when that person is on holiday.

Pain first, platform second

Ansible Automation Platform (and tools like it) are excellent once you know what you’re trying to make reliable. They’re less helpful when the goal is still “we should automate more.”

A better first question is operational: which recurring piece of work costs you attention every week, and who is the single point of failure if it doesn’t run?

That might be user provisioning, certificate renewal, environment rebuilds, or a checklist that lives in someone’s notes. The technology matters less than the pattern: repeated work, tribal knowledge, and a cost you already feel.

Make it runnable by someone else

A script on a laptop is not automation in any organisational sense. It’s still a person.

What changes the game is turning that work into something the team can see and run safely: a job template, clear inputs, credentials that aren’t copied into chat, and a log of what happened. When a colleague can execute the same job without reverse-engineering a shell history, you’ve moved from clever to durable.

That’s also where governance stops being a buzzword. Approvals, inventories, and role-based access aren’t bureaucracy for their own sake — they’re how you let more people help without raising the blast radius.

One boring win beats a portfolio of demos

There’s a temptation to showcase the platform with something impressive. Resist it early on.

One well-owned, boring automation that removes a weekly fire teaches the organisation more than a polished demo of five unfinished ideas. People learn the workflow. Trust builds. The next candidate becomes obvious because someone says, “Can we do the same for this?”

In consulting work, that sequence shows up again and again: find the pain, make it visible, make it runnable by more than one person, then widen. Platforms amplify that loop. They don’t replace it.

A practical starting checklist

• Pick one job you already do at least weekly.
• Write down the inputs, the expected outcome, and what “good” looks like.
• Move credentials and inventory out of personal machines.
• Put the job where a teammate can run it — and where you can see the result.
• Only then look for the second job.

Automation programmes don’t fail because teams lack tools. They stall because the first success was never close enough to real work.

Start with the job that already hurts. Make that one trustworthy. The rest gets easier.