Got pulled into a Slack thread at 7:12 because a terminated contractor was still pushing code.
What made it worse was the ticket already showed completed. Manager approval was there, the offboarding request was closed, and our chat bot had even posted the usual “done” message. Meanwhile the account was still active in source control, VPN was alive, and one cloud group never got cleaned up.
That was the part that finally got me to stop trusting status updates by themselves. We had AI doing the easy part pretty well. It could read the request, figure out it was an offboarding case, summarize what needed to happen, and even tell me which systems were probably involved. But it still wasn’t the thing I wanted making live identity changes on its own, especially when legal hold and manager exceptions show up at the same time.
Before this, our setup was basically a pile of separate automations. One disabled the directory account. Another tried to revoke app access. Another sent the closure note back to the ticket. If one step timed out or hit a weird dependency, the rest could still look finished enough that nobody noticed until someone yelled.
What changed for us with Autom Mate was putting an actual control layer in the middle instead of letting every tool pretend it owned the workflow. Now the AI can suggest the path, but the execution waits on the right approval state, checks for exception flags, and then runs the shutdown steps in order. If the repo access removal fails, the ticket does not get to cosplay as resolved. It stays open, posts the failure back into the thread, and shows exactly which step stalled.
The biggest difference is I don’t have to play detective across five consoles anymore. Offboarding finally behaves like one process instead of a bunch of optimistic API calls. That alone saved us from repeating the same “why is this person still active” conversation the next week.