Hi Everyone,
I'm currently exploring the OOB Conversational Contract Search and Insights Agentic Workflow available in ServiceNow Contract Management Pro and would like to understand its backend execution flow from a technical perspective.
I'm not looking for the functional overview, but rather how the workflow is orchestrated internally.
Specifically, I'm trying to understand the following:
- When a user submits a query, does the workflow first classify it as a structured or unstructured query?
- Is there an intent classification step before selecting a tool or skill?
- How does the agent decide whether to:
- Query ServiceNow tables (GlideRecord/structured search), or
- Use a RAG (Retrieval-Augmented Generation) pipeline to search contract documents?
- Which OOB components are involved in the execution flow?
- AI Agent
- Skills
- Tools
- Prompt Templates
- Flow Designer
- Script Includes
- Is Vector Search or AI Search used internally for retrieving relevant contract content?
- How is the final prompt constructed before being sent to the LLM?
- Does the workflow only generate conversational answers, or can it also update contract metadata?
- Are there any technical design documents, architecture diagrams, or implementation guides explaining this end-to-end backend process?
I'm looking for something similar to the following execution flow:
If anyone has explored this OOB Agentic Workflow or knows of any official ServiceNow documentation, Knowledge sessions, or architecture references that explain its backend implementation, I would greatly appreciate your guidance.
Thank you!
