Incident Management - Auto Resolution based on user input
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
yesterday
When a user submits an Incident via a Record Producer, the system should analyze the user’s input, identify the best-matching keywords, and retrieve the corresponding predefined resolution steps from a custom table. These resolution steps should then be added to the Incident’s Additional comments so the user can see recommended actions and proceed accordingly.
Please suggest a solution approach for this.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
16m ago
Hi sondhal,
1. Recommended approach → Keyword-based lookup (simple & effective)
Instead of heavy NLP/AI, start with:
- Create a custom table (e.g.
u_incident_resolution_map)
Fields: - Keywords (text)
- Resolution steps (HTML/Text)
- Category/Subcategory (optional for better accuracy)
2. Capture user input from Record Producer
From Record Producer:
- Use fields like:
- Short description
- Description
Pass these values to the Incident (standard behavior)
3. Match keywords (core logic)
Use a Business Rule (after insert) OR Flow Designer
Logic:
- Take user input (description)
- Compare with keywords table
Example (simple approach):
var gr = new GlideRecord('u_incident_resolution_map');
gr.query();
while (gr.next()) {
if (current.description.toLowerCase().includes(gr.u_keywords.toLowerCase())) {
current.comments = "Suggested Resolution:\n" + gr.u_resolution_steps;
break;
}
}
You can enhance with:
- Multiple matches
- Priority ranking
4. Populate Additional Comments
- Use:
comments(Additional comments)
So user sees suggestion immediately after creation
5. Optional → Improve accuracy
You can enhance later with:
- Category-based filtering
- Multiple keyword matching
- Scoring logic
6. Alternative (if using Virtual Agent / AI)
If your instance has:
- Virtual Agent
- Predictive Intelligence
Then you can:
- Use similarity matching / classification
But for most cases:
Keyword-based solution works well and is easy to maintain
