Operations · · 4 min read
Teach the daily report when to stay quiet
When should a scheduled agent report send an update, and when should it say nothing?
At nine every morning, an agent posts the same project summary. Nothing has changed, but the channel gets another long message. A week later, an actual blocker appears halfway down the familiar text and nobody notices. The agent kept its schedule and weakened the team's ability to see exceptions.
An Atlassian Community question asks how to suppress a daily email when its query finds no matching issues. For an agent, no meaningful change, no matching work, and a failed check are three different outcomes.
Decide what deserves a person's attention#
Start from the recipient's decision. A project lead needs to know when a milestone becomes late, an owner changes, or a customer dependency is resolved, not a paraphrase of every unchanged task. A compliance report or contractual daily notice can be required even when nothing changed; record that explicitly.
Take an agent that checks customer onboarding blockers each weekday:
- It notifies the account owner when a blocker appears, worsens, clears, or reaches an escalation point.
- It stays quiet while the same known blocker remains within the agreed response window.
- If the source connection fails, it alerts the workflow owner that the check was incomplete.
Write the notification rule so a manager can inspect it. "Notify for a new blocked account, changed blocker owner, missed target date, or restored readiness" is testable; "send anything important" leaves the decision to a model without enough business context.
Compare facts before composing prose#
The agent compares the current business state with the last successfully checked state. Comparing the wording of two generated summaries creates false changes, because wording varies when facts do not.
| Check result | Business meaning | Intended behavior |
|---|---|---|
| Same relevant facts | No new decision needed | Stay quiet and record the check |
| New or changed blocker | A person needs to act | Send the change and next action |
| Known blocker passes escalation time | Delay is now meaningful | Escalate once under the rule |
| Source cannot be read | Monitoring coverage is incomplete | Notify the workflow owner |
| Blocker resolved | People can resume dependent work | Send the resolution if useful |
Retain the comparison point only after a successful check. If a partial read becomes the baseline, missing records look like resolved problems; a failure should not erase the last reliable view.
Keep reminders separate from changes#
Some work needs a reminder even when the facts are unchanged. Define its cadence separately: who gets reminded, after what interval, and when it escalates. Otherwise the daily report becomes an accidental reminder system.
When someone acknowledges a blocker, preserve that and stop repeating the introduction. A later reminder says why attention is needed now, such as a deadline approaching without an accepted next step.
Keep a notification short enough to act on: what changed, which account or task, the source, and who owns the next action. Include the prior value when it explains the change; "target moved from Tuesday to Friday" beats "timeline updated."
Test silence as a successful result#
Run the schedule twice against unchanged data. The second check should stay quiet and write no new report. Then test a new blocker, a cleared blocker, a source failure, and an overdue unchanged blocker.
For duplicate alerts, inspect the saved comparison point and notification history. For a missed change, inspect whether the source was fully read and the rule covered that field. Forceful prompting does not fix a baseline that never persisted.
Review the rules with recipients after real examples; remove noise, add missed transitions.
Also check the routine after several missed schedules. It should compare against the last reliable state and report the accumulated change once; replaying every missed daily notification floods a channel with obsolete updates. Keep missed-check history for the workflow owner and give the business recipient the current condition and any open decision.
Use your agent with Deixic routines and company methods for recurring work, and inspect action history when a check or delivery fails. A quiet day is successful when the check happened and the team can tell silence from lost coverage.