AI notes for Troubleshooting
AI notes are structured records generated by ServiceMax Core from completed Work Orders. The Troubleshooting Agent uses them as its knowledge base when diagnosing active service issues.
Overview
Based on the Troubleshooting Agent configuration, the Service Intelligence Data Pipeline (SIDP) scheduler runs and automatically processes completed Work Orders and generates AI notes. Each AI note captures the problem that occurred and the resolution applied, giving the Troubleshooting Agent a searchable record of past service activity. Over time, these notes form a knowledge base that the agent draws on to surface relevant insights during active troubleshooting.
By default, AI note generation runs on a schedule for 24 hours. As Work Orders are completed and fall within the configured data scope, the SIDP scheduler processes them and adds their notes to the knowledge base.
Considerations
AI note generation requires the Troubleshooting Agent to be configured with at least one Problem Overview data source field and the required Context Data fields before processing begins. Work Orders that fall outside the configured data scope are not processed. The Solution Overview configuration cannot be completed until the Problem Overview configuration is saved. For more information, see
Managing Troubleshooting Agent Configurations.
How it Works
When a completed Work Order is processed, the SIDP scheduler retrieves data from Salesforce using the instructions defined in the Troubleshooting Agent configuration. It maps that data to AI note fields according to the configured data mappings, then generates two AI-written summaries: a Problem Overview and a Solution Overview. The AI notes are stored in Snowflake.
The Troubleshooting Agent performs a semantic search to find AI notes that are contextually similar to a current issue. Keyword matches alone are insufficient; the semantic approach surfaces relevant past cases even when the language differs. If a Work Order contains missing or incoherent data, the system flags it and excludes it from the knowledge base.
At runtime, when a technician runs a troubleshooting AI action on an open Work Order, the system generates a temporary AI note that contains only a Problem Overview. The Troubleshooting Agent uses this AI note as the query input for semantic search against the stored knowledge base and provides similar troubleshooting summaries.
Key Capabilities
• Problem Overview: AI-generated summary of the problem that needed to be solved, including relevant context such as attempted solutions and similar past issues on the asset.
• Solution Overview: AI-generated summary of the resolution and how it was implemented, including information about parts and tools used.
• Structured Metadata: Each AI note stores Asset Information (the details needed to understand the asset context) and Work Order Information (date performed, duration, customer, and technicians involved) as structured fields, separate from the AI-generated summaries.
• Record references: Unstructured summaries include structured references to specific records, such as Product records for parts mentioned in the text. These references follow the same format used in AI Action and AI Chat responses, supporting clickable record names in the user interface.
• Bad data detection: The system flags Work Orders with missing or incoherent data and excludes them from the knowledge base automatically.
Related Topics