Skip to content

Automation & Orchestration

Notification automation: the right person, once, with what they need

A late technician, a failed payment, a gas report: each one matters to someone specific, and each deserves a different kind of interruption. Notification automation decides from the event itself who owns it, how urgent it is, which channel to use, and who's next if nobody responds.
Discuss your systemSee the policy ↓
EventsPolicyWho hearsjob.running_lateJ-512 · T2job.running_lateJ-512 · againpayment.failedINV-2231gas.reported14 Elm StPolicyowner · urgencychannel2 events · 1 message25 min late; customer toldOffice leadin-appadded to today's digestFinance17:00 digestgas report, 14 Elm StOn-call T5SMS nowcalling nowOperations managerbackupno reply · 10 min

Too many alerts is the same as none

Notifications fail in two ways. Nobody hears about the thing that mattered, or everyone hears about everything and stops reading. A shared channel full of automated messages usually manages both.

Notification automation starts from the event and its owner. It picks one person by role and rota, a channel that matches the urgency, a message that says what to do, and a rule for what happens if nobody responds. Everything else waits for a digest or is simply recorded.

The policy, written down

Every notification comes from a rule like these. You can read them, and change them, without touching a workflow.
Technician running more than 15 minutes lateOffice lead, in the job app; the customer is told automaticallyNothing; it's for information
Payment failedFinance, in the 17:00 digestIncluded again the next day, marked overdue
Gas, smoke or flooding reportedOn-call technician by SMS, immediatelyOperations manager called after 10 minutes
Quote approved by a customerThe salesperson who sent it, by emailOffice lead after one working day
A workflow failedThe workflow's owner, in the operations channelRepeated hourly until acknowledged

What a good notification carries

  1. 01

    Who it's for

    One owner

    A person chosen by role and rota, not “everyone in the channel”.

  2. 02

    Why now

    The event

    What happened, in one sentence, with the job or customer it's about.

  3. 03

    What to do

    The action

    The decision or task, with a link straight to it.

  4. 04

    By when

    The deadline

    How long they have, and what happens after that.

  5. 05

    What's already done

    The context

    Steps the system or a colleague already took, so nobody repeats them.

Now, later, or not at all

  1. Instant

    Someone must act soon

    SMS, a push or a call, for the few events that can't wait. Escalates if unacknowledged.

  2. Digest

    Someone should know today

    Collected into one message at a set time, so routine events don't interrupt anyone.

  3. Record only

    Someone may need it later

    Kept on the job's history, searchable, with no message at all.

What notification automation hands off

In

Events
from your systems
Ownership
roles and on-call rota
Rules
urgency, channel, quiet hours
Notification policy
who · how · when · then what

Out

Messages
one owner, one channel
Escalations
when nobody responds
Log
who was told, who acknowledged

Worth building when the important alerts get lost

Good fit

  • Events in your systems need a person to act, and the right person depends on the event.
  • Some events are urgent and some aren't, but today they all arrive the same way.
  • Nobody can say who was told about what.

Not the right fit

  • A handful of events a week that one person already handles.
  • The ownership of the work behind the alerts hasn't been agreed.

Where it's built in

Next to notification automation

Questions about notification automation

Which channels can it use?

The ones your team already uses: SMS, email, the app they work in, chat tools such as Slack or Teams, and a phone call for the urgent few.

How do you stop alert fatigue?

Fewer, better notifications: related events merged, one owner instead of a channel, digests for anything that isn't urgent, and a regular look at which alerts nobody acts on.

What about evenings and weekends?

Quiet hours per person and per channel, with an on-call rota for the events that can't wait until morning.

What if someone ignores it?

Escalation is part of the rule. After the time you set, the backup is told, and the log shows who was notified and when.