It is the year 2010. Your team finishes a feature on a Friday. The release manager schedules the deploy for the third Wednesday of next month, because that is the change window. Twenty other features are queued for the same window.
On the night of the deploy, half the team stays late. Somebody runs a script. Something goes wrong, and nobody can tell whose change caused it — there are twenty changes. The rollback plan is "restore last night’s backup". By 3 a.m. the site is back up and nobody wants to do this again for a while.
A CI/CD pipeline kills this story. Every commit is built, tested, scanned and deployed on its own. Twenty small changes become twenty small deploys. If one breaks, you know which one — and rolling it back is one click.