- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-20-2021 06:10 PM
Dear all,
I have seen a very cool demo about virtual agent done by servicenow:
When you type any request at the beginning of the virtual agent(such as business travel), It can automatically recognize whether you are looking for an related article, an item, or an incident(It can even identify that you need live agent) and return the result to you. I want to know how to achieve it.
Now I only know that you can identify topics through the NLU module, but this is far from enough.
Can someone tell me how to implement this advanced function? Thank you.
Solved! Go to Solution.
- Labels:
-
Agent Chat
-
AI Search
-
Virtual Agent

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-21-2021 12:42 PM
Besides NLU and having a solid foundation of topics, I think you want/need the (AI) Search Fallback setup topic as well. By default AI Search searches Knowledge and the Catalog - and from my usage it cannot index and therefore search tasks (INCs/RITMs/etc) at least when I recently tried in Quebec. I suppose if not using AI Search though, you could use a Search Context that includes Incidents and more (though I'd question the value in that and probably advise against it). This is assuming you meant "looking for... an incident (record)". If you meant "looking for... (creating) an incident", disregard what I just said - that's just basic VA functionality you can do by creating topic(s) that create incidents.
With the Search Fallback setup, if VA can't match the user's utterance to a topic, then it'll present them with relevant catalog items/knowledge articles. If that can't help them, I would recommend - this is what we have - allowing them to create an incident or do a live chat as part of the fallback topic. That should cover the entire range of what you want - and it's what we aspire to (we just need to setup AI Search first).
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-21-2021 05:09 AM
This would all be done with entity recognition.
You'd build entities for ticket numbers and catalog items and then tie those to an intent for each. If neither matches then you'd fall back to looking for an article.
Check out the VA Academy video on entity recognition:
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
07-22-2024 07:18 AM
can you provide little more details on how to use it to show incident cards

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-21-2021 12:42 PM
Besides NLU and having a solid foundation of topics, I think you want/need the (AI) Search Fallback setup topic as well. By default AI Search searches Knowledge and the Catalog - and from my usage it cannot index and therefore search tasks (INCs/RITMs/etc) at least when I recently tried in Quebec. I suppose if not using AI Search though, you could use a Search Context that includes Incidents and more (though I'd question the value in that and probably advise against it). This is assuming you meant "looking for... an incident (record)". If you meant "looking for... (creating) an incident", disregard what I just said - that's just basic VA functionality you can do by creating topic(s) that create incidents.
With the Search Fallback setup, if VA can't match the user's utterance to a topic, then it'll present them with relevant catalog items/knowledge articles. If that can't help them, I would recommend - this is what we have - allowing them to create an incident or do a live chat as part of the fallback topic. That should cover the entire range of what you want - and it's what we aspire to (we just need to setup AI Search first).
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-21-2021 07:22 PM
Thank you very much for your detailed explanation.
By the way, can you describe the difference between AI search (fallback) and Contextual search (fallback)? I found them in the topic list, but I don't know when to use which one?
Thanks a lot.