Objective
This guide walks you through building an agentic complaint triage workflow in ServiceNow from start to finish. By the end, you will have a working Agentic Workflow that classifies inbound customer complaints, researches relevant context from a knowledge base, and guides supervised case creation and routing, using the same lab built for Knowledge 2026 session LAB8119.
Before You Start
This guide assumes you already have a ServiceNow instance with the lab assets deployed. If you have not done that yet, start by confirming these features are active on your instance:
Activate/Configure these features first
- FSO Complaint Management must be installed. Refer to the FSO Docs for setup instructions.
- AI Agent Studio must be installed. See the AI Agent Studio docs for details.
- Now Assist Panel must be activated. Follow the activation guide if you have not done so.
Upload these assets from the repository
Head to the K26-LAB8119 GitHub repository and upload the following to your instance:
- Agents — Prebuilt AI agents that classify customer interactions and support outbound communication
- Interactions — Simulated inbound customer emails that serve as your test data throughout the exercises
- KB — Knowledge Base articles the agents reference when researching complaint context
Note: All three asset sets need to be loaded before you execute any of the steps below. The lab guide will not work as expected without them.
Once your instance is ready, you can follow along using the full step-by-step guide on GitBook.
What You Will Build
The Triage Agentic Workflow
The core of the lab is a multi-step agentic workflow that handles inbound e-mails with minimal human intervention until the moment a case needs to be approved and assigned.
Intent Classification
The first agent reads the inbound interaction and classifies the complaint type. This is where the preloaded Agents folder comes in, the classification agent is already configured with the right prompts and tools. Your job is to wire it into the workflow and validate that it fires correctly against the simulated Interaction records.
Knowledge Base Research
Once the complaint is classified, the agent searches the preloaded KB articles for relevant policy and procedural context. This step is what allows the agent to synthesize documented policies and procedures that an operations associate can actually act on, rather than just a raw email dump.
Supervised Case Creation and Routing
The final step in the workflow creates a case record and routes it to the appropriate queue based on the classification output. This is a supervised action, a human reviews and approves before the case is committed, which keeps the workflow compliant with FSO operational standards.
Key Concepts
- AI Agent Studio
Is the ServiceNow environment where you design, configure, and test AI agents. Think of it as the authoring surface for agentic workflows. - Agentic workflow
Is a sequence of AI-driven steps that can plan, act, and adapt based on input, unlike a traditional flow that follows a fixed path regardless of content. - Interactions
In ServiceNow represent inbound communications, in this case simulated customer emails. They are the entry point that kicks off the triage workflow. - Supervised actions
Are steps where the AI prepares an output but holds for human review before committing. This pattern is essential in regulated industries where auditability matters.
What's Next
If you want to go deeper after completing the lab, a few natural next steps from here:
- Explore how the classification agent handles edge cases by modifying one of the Interaction records and re-running the workflow
- Review the FSO Financial Services Operations community for discussions on AI-powered complaint handling
- Check out the ServiceNow AI Agent Studio documentation for the full configuration reference beyond what this lab covers
