Celonis Smart Sensor
The Celonis Smart Sensor is an automated monitoring engine that scans your data models in real-time to detect specific business incidents, known as Signals. By bridging the gap between data refreshes and actionable insights, it ensures that your business logic—defined through PQL filters—automatically triggers the right workflows the moment a data anomaly or requirement is identified.
The choice to use the manual or smart sensor depends on the following factors:
Factor | Manual sensor | Smart sensor |
|---|---|---|
Trigger source | Human action (URL click/Run button). | Data event (PQL query match). |
Input data | Provided by the user at runtime. | Pulled automatically from the Data Model. |
Control | High; user decides exactly when to run. | Automatic; runs whenever conditions are met. |
Primary use case | Ad-hoc tasks, testing, or human approvals. | Continuous monitoring and high-volume automation. |
Visibility | Requires the user to have the URL. | Runs in the background. |
Before you begin, ensure you have a published Data Model and a Knowledge Model (KM) containing the records you wish to monitor.
The Smart Sensor acts as the bridge between your data and automated action. Follow these steps to define which records should trigger a Signal and how those Signals are maintained.
Select the Knowledge Model: Choose the KM that contains the business logic and records the sensor will scan.
Identify the Record: Select the specific Record type (e.g., Invoices, Sales Orders) that the sensor will monitor.
Define Signal Filters: Add one or more PQL filters to isolate specific business incidents.
Tip
Multiple filters are joined by AND logic; a Signal is only created if every condition is met.
Set Sensor State: Toggle the state to Active to allow the sensor to run automatically in the background.
Save and deploy: Click Save, then deploy the Skill.
Because filters are stored directly in the Knowledge Model, you must deploy both the Skill and the respective Knowledge Model. If you only deploy the Skill, your logic changes will not take effect.
When adding filters, you have two primary options within the configuration menu:
Create New Filter: Opens a PQL editor.
Requirement: Every statement must start with the
FILTERkeyword.Naming: The Filter ID must only use alphanumeric characters, dashes (-), or underscores (_).
Select Existing Filter: Choose logic already defined in your Knowledge Model. This is recommended to ensure a "single source of truth" across multiple Skills or Views.
Use these integrated tools to verify your setup before it goes live:
Signal Preview Table: Click Show Signal Preview to see a real-time list of records meeting your criteria. This allows you to verify your PQL logic and column sorting without executing the entire Skill.
Sensor Logs: Once published, check the Sensor Logs section within the Skill Logs to audit executions and track newly created or resolved Signals.
Capacity: Each execution can process a maximum of 10,000 signals. If more are found, they are processed in a queue based on their record identifier.
Automatic Triggers: The sensor re-scans for Signals whenever:
The Data Model is reloaded.
The Knowledge Model is deployed.
The Skill is deployed.