We used to call things 'the GPT one' and 'the new one'. Now every agent is named after its job: Report Drafter, Inbox Router, Notes Cleaner. When we swap the model underneath, nobody has to relearn anything.
Small thing, but it ended a lot of confused Slack threads. Curious whether others name them, or whether that is overthinking it.
Same here, and we add the owner's name in the description. If the Inbox Router does something strange, everyone knows who to ask.
We do this internally too. The job name also makes the change log readable months later.
We do, and we version them too: Readout Drafter v3. When the readouts change tone, the version number tells you why.
Calendar Reader, Brief Writer, Draft Checker. The names ended the 'which bot did this' question in our editorial channel.
Same, with one more rule: the name has to fit in the automation tool's node label. Forces you to describe the job in two words.
Coming from design systems this feels very familiar. Name by purpose, not by implementation. Stealing the owner-in-description idea.