Escalation Rules

Escalation rules guarantee that no critical problem goes unattended. When an alarm is not acknowledged, when maintenance has been overdue for days, or when an urgent task has not moved in hours, the system escalates automatically — first notifying the technician, then the supervisor, then the manager — until someone responds.

What is it for?

In plants that operate 24/7, the greatest threat is not detecting problems — it is that no one responds in time. A critical alarm that appears at 3 AM can remain active until dawn if the shift technician was somewhere else or missed the notification.

Escalation solves this with an automatic notification chain:

• If the technician does not respond within X minutes, the supervisor is notified
• If the supervisor does not respond within Y minutes, the plant manager is notified
• If the manager does not respond, escalation can continue to the operations director

This creates a safety net ensuring every critical problem always reaches someone with authority to act.

How does it work?

Escalation rules are evaluated automatically every minute. The system checks:

• Are there unacknowledged alarms that have been active longer than the defined threshold?
• Are there scheduled maintenance tasks that have been overdue longer than the limit?
• Are there urgent tasks that have passed their deadline?

When a condition is met, the system executes the notification steps in order. Each step has a cumulative wait time (delay) from the moment of the original event.

How to use it?

Create an escalation rule

1. Go to Notifications > Escalation in the sidebar.
2. Click New Rule.
3. Define the trigger type:

4. Define conditions to filter when the rule applies:

5. Define the escalation steps:

Each step has:

• Label: description of the step (e.g., "Notify on-call technician")
• Delay in minutes: time from the original event before executing this step
• Notification channels: email, WhatsApp, and/or in-app notification

Example escalation chain for a critical alarm:

6. Enable the rule with the Enabled toggle.
7. Save.

How escalation stops

Escalation stops automatically when:

• The alarm is acknowledged by someone
• The task is completed or cancelled
• The maintenance is executed

Once someone responds, the pending steps are not executed.

Difference between agent escalation and rule escalation

There are two ways to configure escalation in Rela AI:

In the Machine Agent (for alarm-specific agent escalation): Configured directly in the agent. Applies to all alarms processed by that agent. Simpler for cases where a single agent manages a zone or piece of equipment.

Escalation Rules (in this section): More flexible — can apply to any type of event (alarms, overdue tasks, maintenance). Allows more detailed conditions and is independent of each agent's configuration.

Escalation audit trail

Each escalation step execution is recorded with:

• Which rule was executed
• Which step in the chain
• Who was notified
• Whether the notification was delivered successfully or failed
• Exact timestamp

This record allows answering questions like: "Did the critical alarm on Saturday reach the manager?" and getting the exact answer with evidence.

Key benefits

• No critical alarm goes unanswered even if the technician misses the initial notification
• Configurable chains with multiple steps and channels (WhatsApp, email, in-app)
• Detailed conditions to escalate only what truly matters
• Stops automatically when someone addresses the problem
• Complete record of all escalations for compliance and review
• Applies to overdue maintenance and delayed tasks, not just alarms

Common use cases

Scenario 1: Critical alarm on the night shift The "Critical Alarm No Response" rule is configured for: any alarm of Critical or Emergency severity from industrial sources. Step 1 (5 min): WhatsApp to the on-call technician. Step 2 (15 min): Email + WhatsApp to the shift supervisor. Step 3 (30 min): Email to the plant manager.

At 2:17 AM, compressor C-02 generates a critical alarm. The technician is doing a round and does not see the WhatsApp immediately. At 2:22 AM (5 minutes later), the system sends the WhatsApp. The technician sees it, goes to the compressor, and acknowledges the alarm at 2:24 AM. Step 2 (supervisor) is never executed because the technician responded in time.

Scenario 2: Overdue preventive maintenance The "Overdue PM" rule is configured for maintenance with more than 3 days of delay. Step 1 (3 days overdue): Email to the maintenance coordinator. Step 2 (5 days overdue): Email + WhatsApp to the maintenance manager. Step 3 (7 days overdue): Email to the plant manager.

The monthly PM for generator GEN-01 was not executed on the scheduled date. On the third day of delay, the coordinator receives an email alert. If they do not schedule it within the next 2 days, the maintenance manager also receives the alert.

Scenario 3: Stuck urgent task The "Stalled Urgent Task" rule applies to Urgent priority tasks with more than 2 hours without a status change. Step 1 (2 hours): In-app notification to the supervisor. Step 2 (4 hours): WhatsApp + email to the maintenance manager.

An urgent repair task is created at 10 AM but no one moves it to "In Progress." At 12 PM, the supervisor receives the in-app alert. At 2 PM, if it still has not moved, the maintenance manager receives the WhatsApp.