Pipelines that let you deploy on a Friday, with a rollback somebody has actually run rather than one that exists on a wiki page.
A long, error-prone deployment is a symptom of missing automation rather than of team culture. The team already knows what the manual steps are. They have written them down, usually in a document that is out of date, and they take turns doing them on a Tuesday because Tuesday leaves four days to notice.
The work is turning those steps into a pipeline that runs the same way every time, and making the rollback a thing somebody has executed rather than a paragraph nobody has read.
ICF is early, so there are no named client logos here yet. What there is: public code you can open and read, and work described without naming whose it was.
01
Read the pipeline
What runs today, how long each stage takes, and which stages people skip when they are in a hurry. The skipped ones are the interesting ones.
02
Agree the target
What "deployable on a Friday" means for your system, written down, including what you are willing to give up to get there.
03
Rebuild in slices
One pipeline at a time, in front of the team, so nobody inherits something they have not seen built.
04
Rehearse the rollback
A rollback that has never been run is a paragraph. We run it, on purpose, before you need it.
Deployments happen on a Tuesday so there is time to notice
The rollback plan is a document rather than a command
Each repository has its own copy of the same pipeline, drifting
A release needs somebody specific to be awake
Tell me what you are running and what is slowing you down. I reply within one business day, and I will say so if this is not work I should take.
Taking a small number of contracts