Building a Notification Bot Turned Me Into Its Support Desk
A note on how small notification bots blur the lines around accuracy and responsibility as they become more useful.
Notification bots are useful when kept small. Simply posting daily counts, open-item lists, and overdue tasks to Slack saves the trouble of checking dashboards.
But the more useful they become, the less their output stays “for reference.” Team members act on bot notifications. If nothing arrives, they assume there are no issues. When a false alert or missed notification occurs, accountability suddenly becomes unclear.
The builder tends to think: if the input data is correct, the notification will be correct. The user, on the other hand, assumes that because it came from the bot, it must be right. This gap in perception means a small bot can have an outsized operational impact.
Recently, I have started including the following in every notification:
Scope of aggregation
Aggregation timestamp
Link to the source data
Contact for anomalies
Brief description of notification conditionsEven this much makes it easier for the reader to verify what the notification is based on.
I think the same applies to AI agent output. Once it starts being used, it becomes part of operational decisions, not just a suggestion. That is why the design needs to show not only the output itself, but also the rationale, the scope, and where to go when something goes wrong.
