Skip to main content

Celonis Machine Learning Sensor

The Celonis Machine Learning Sensor allows you to automate responses to complex data patterns by integrating advanced Python algorithms and machine learning models directly into your workflows. Unlike standard sensors that monitor simple thresholds, the ML Sensor identifies "Signals"—specific data incidents defined by sophisticated logic—to trigger automated Skills.

Integrating machine learning into your automation strategy provides several key advantages:

  • Advanced Pattern Recognition: Detect complex data incidents that standard PQL or manual rules might miss, such as duplicate invoice patterns or predictive delays.

  • Real-Time Responsiveness: The sensor automatically scans your Data Model for new Signals every time the model is reloaded or when KM/Skill changes are published.

  • Tailored Use Cases: Leverage built-in ML templates for common scenarios or configure Custom ML Sensors to meet specific business requirements.

  • Proactive Operations: Shift from reactive data monitoring to proactive intervention by triggering immediate actions (Tasks, Webhooks, or updates) as soon as a machine learning model identifies a risk or opportunity.

Configuring a Machine Learning Sensor

Follow these steps to set up a Machine Learning Sensor within your Celonis Studio environment:

  1. Open your Skill: Navigate to Studio and open the specific Skill where you want to add the trigger.

  2. Select the Sensor: In the Skill editor, click Add Trigger (or the + icon) and select Machine Learning Sensor from the list of available sensors.

  3. Define the Knowledge Model: In the sensor settings, select the Knowledge Model (KM) that contains the data objects you wish to monitor.

  4. Choose the Use Case: Select a Use Case from the dropdown menu. You can choose from:

    • Pre-defined Use Cases: Built-in ML logic provided by Celonis.

    • Custom ML Sensor: Select this to apply your own advanced Python algorithms.

  5. Configure Criteria: Define the specific conditions or thresholds the ML model should look for to identify a "Signal."

  6. Save and Deploy: Click Save and then Deploy your Skill. The sensor will now begin scanning your Data Model for Signals based on the reload schedule or model updates.

Tip

Ensure the Data Model referenced in your Knowledge Model is correctly mapped; the sensor relies on these data reloads to identify and act on new incidents.

Related topics