CSM AI Agent: Troubleshooting Steps Identification

[FernandoCastro]

The Troubleshooting Steps Identification AI Agent analyzes the content of a case, identifies missing or incomplete context, compares it against relevant knowledge, and recommends the next best troubleshooting actions for live agents. It reduces the time agents spend searching for information by surfacing the most relevant steps directly in the flow of work.

Key Capabilities

• Extracts problem details from case fields, email content, or past interactions.

• Identifies gaps or missing context and highlights what data must be collected.

• Searches across knowledge and related content to match the case to appropriate troubleshooting procedures.

• Provides structured, actionable next steps that agents can follow immediately.

• Integrates with retrieval tools to surface the most relevant content for diagnosis.

Implementation

1. Install Now Assist for CSM

2. Open AI Agent Studio and select the Troubleshooting Steps Identification agent.

3. Add a retrieval tool and configure it to target the knowledge sources that contain troubleshooting procedures.

4. Define the trigger for when the agent runs, such as when a case is opened or enters triage.

5. Validate the agent using sample cases to confirm the recommended steps are accurate and grounded.

6. Monitor usage and refine retrieval filters or knowledge sources as needed.

More information here:

• Using AI Agent
• Configuring AI Agent

Key Best Practices

• Ensure knowledge articles and troubleshooting guides are accurate, current, and structured consistently.

• Apply filters to retrieval tools so only high-quality content is used for recommendations.

• Start with a small set of case types to maintain precision before expanding coverage.

• Gather ongoing feedback from agents to refine the quality of recommended steps.

Formal Learning

• Now Learning courses for Now Assist for CSM

• AI Agents Delivery Accreditation

• Now Assist AI Agents Deep Dive

Measured Success

Outcome	Explanation	Success Metric
Faster issue diagnosis	Agents receive structured steps without manual searching	Time to first troubleshooting action
Higher relevance of recommendations	Steps align with case context and knowledge	Percentage of steps accepted
Improved consistency across agents	Standardized troubleshooting procedures are followed	Variability in steps taken
Reduced search time	Less manual navigation of knowledge content	Lookup time per case

FAQs

How does the AI Agent determine which steps to recommend?
It compares case content with indexed knowledge sources to identify the best matching troubleshooting actions and highlights missing information as needed.

Can I customize which knowledge sources the agent uses?
Yes. Admins can configure retrieval tools to search only specific knowledge bases, categories, or tagged content.

Does this work for all case types?
It is most effective when troubleshooting procedures are documented and follow consistent patterns. Start with high-volume or repeatable issues.

Can agents override or skip recommendations?
Yes. Recommendations support agents but do not enforce a required workflow. Agents can follow, adjust, or bypass suggestions based on the situation.

How do I improve the accuracy of recommendations?
Keep knowledge content well-maintained, remove outdated articles from retrieval scope, and refine filters based on live agent feedback.