Provide actionable next steps by leveraging patterns from past cases
Suggested Steps Generation recommends the most effective next actions for customer service agents based on how similar cases have been resolved historically that may improve vastly the customer experience. The generative AI skill analyzes case clusters to offer guidance that helps agents resolve issues more quickly and consistently across the organization. See below in the Recommendations, right side panel how these are displayed:
Key Capabilities
-
Smart Next-Step Suggestions
Agents see tailored suggested actions (such as knowledge articles or past Case solutions) in the workspace based on the currently open case, account and driven by similar past Case resolutions. -
Powered by GAF Clustering
The Group Action Framework identifies clusters of resolved cases and generates action guidance accordingly, tailored to case attributes and outcomes. -
Embedded in Case Workspace
Suggestions appear within the CSM Configurable Workspace side panel for easy access during resolution. -
Built-in Feedback Interface
Agents can rate suggestions positively or negatively, providing continuous feedback to improve the quality of future recommendations.
Implementation
Prerequisites:
-
Now Assist for CSM plugin enabled
-
AI Search and Group Action Framework (GAF) configured and active
-
Clustering configured using existing high-volume resolved case groups/use cases
-
Case resolution taxonomy structured consistently using case management best practices
Configure Suggested Steps Generation (Product Docs link coming soon)
Use Suggested Steps Generation
Key Best Practices
-
Use consistent taxonomy and resolution codes to improve clustering reliability
-
Train agents to give quick thumbs up/down feedback to improve recommendations over time
-
Monitor usage and accuracy with dashboards from the Now Assist Admin Console
- Setup GAF
- Generate synthetic data to meet required cases for clustering
Formal Learning
Now Learning: Now Assist for CSM Essentials Path
Measured Success and Outcomes
|
Outcome |
How It Supports Customer Service Teams |
Success Metric |
|---|---|---|
|
Time saved on common cases |
Agents apply pre-validated steps instead of recreating from memory |
Time to resolution for repeated case types |
|
Greater consistency |
Standardized approach to similar issues minimizes variability in handling |
% of cases using suggested steps |
|
Reduced training overhead |
New agents can follow proven steps without deep process knowledge |
Time to proficiency for new agents |
|
Confidence in resolution |
Clear guidance reduces hesitation and unnecessary escalations |
Agent confidence and escalation rate |
|
Continuous content improvement |
Feedback loops help refine suggested actions over time |
Improvement in feedback score over time |
Frequently Asked Questions
Where are the suggestions displayed?
Within the CSM Configurable Workspace, in a Suggested Steps side panel triggered when clustering finds relevant similar cases or Now Assist Panel.
Are agents required to follow suggestions?
No. Agents can edit or dismiss suggestions and retain full control over resolution decisions.
Which types of cases work best with this feature?
High-volume, repetitive cases such as shipping issues, complaint issues or login errors perform best as clustering improves with structured resolution patterns.
How frequently are suggestions updated?
GAF uses a scheduled process to update clusters and this is part of the configuration in Now Assist Admin console.
Does this feature work in self-service portals or UI16?
No. Suggested Steps Generation is available only in the CSM Configurable Workspace for customer service agents.
Can usage of this feature be restricted to roles or departments?
Yes. Use the Now Assist Admin console to configure access based on roles, case type, or resolution conditions.
What is the minimum of cases this skill needs in order to work correctly?
The clustering process requires a minimum of 2000 good quality cases in your instance, for optimal results attempt to have a baseline of 3000.
