Self-Service Catalog Suggestion Agent
Recommends the most relevant self-service catalog item or help article to customers before a ticket is created, based on their intent.
Customers frequently submit tickets for issues that already have a fast, self-service resolution path available in the service catalog, but they never discover it because the catalog is hard to search or buried in navigation
Service catalogs themselves often grow large and disorganized, with overlapping items, outdated entries, and inconsistent naming that makes it difficult even for a well-intentioned customer to find the right item
Support teams rarely have visibility into which catalog items would have prevented a ticket, so the catalog does not improve based on real deflection opportunity
Generic search bars using simple keyword matching frequently fail to interpret a customer's actual intent, especially when the customer describes their problem in different language than the catalog item's title
The agent interprets the customer's stated issue or search query at the moment of intent, whether in a help widget, a ticket submission form, or a chat opener, and matches it against the full service catalog using semantic understanding rather than exact keyword matching. It presents the most relevant self-service items ranked by likelihood of resolving the issue, and if the customer proceeds to file a ticket anyway, it logs which catalog items were shown and why they were not sufficient. This deflection and rejection data is used to continuously refine both the ranking model and the catalog content itself.
Intent Interpretation
- Parse the customer's query or ticket subject for intent
- Normalize informal language against catalog terminology
- Identify the underlying issue category
- Detect urgency or complexity signals
Catalog Matching
- Search the full self-service catalog using semantic matching
- Rank candidate items by predicted resolution likelihood
- Filter out outdated or low-quality catalog entries
- Present the top matches at the point of intent
Deflection Tracking
- Log whether the customer resolved via self-service or proceeded to file a ticket
- Capture reasons the suggested item was insufficient
- Measure deflection rate by catalog item and issue category
- Flag high-ticket-volume gaps with no matching catalog item
Catalog Improvement Loop
- Recommend new catalog items for high-volume unmatched issues
- Flag outdated or duplicate items for cleanup
- Adjust ranking model based on deflection outcomes
- Report deflection savings to support leadership