Conversation type FAQ or Q&A?

Mikael Karlsson
Kilo Sage

Hi. Will soon be migrating from several products into ServiceNow.

Our current chatbot will answer a question directly "conversation style" if the match score is good enough and provide topic/intent choice if uncertain. How could this be implemented in the Virtual Agent? Can i have a fallback which is a table with Question -> Answer where if the utterance matches the question good enough causes the VA to spit out the answer as a reply? 

1 ACCEPTED SOLUTION

Chris D
Kilo Sage
Kilo Sage

This I think is arguably one of the biggest weaknesses of ServiceNow's VA compared to others (to be honest, I'm thinking of one in particular I've seen but don't recall the name).

You can certainly build it into VA, but it'll require a bit of design and implementation effort.

As you mentioned, the fallback topic would be the place to start, but something to consider is if you're using AI Search, the AI Search Fallback does come first and that may or may not be helpful for you. In my opinion, using your exact use case, the ootb AI Search Fallback may not be too great because KB results don't show the content - just the Short Description and a real brief blurb with a link to the full article. Genius Results in theory could be useful to show (most) of the (short) article's content, but I still find Genius Results too unreliable and unpredictable.

There is a nice FAQ Conversation Builder topic block which basically does what you want (displays the entirety of the content in-line - not a hyperlink - from FAQ KB articles), but it is limited to FAQ articles specifically and you'd have to plug it into a custom (AI Search) Fallback topic using the user's utterance (vaSystem.searchText iirc) as the search input.

View solution in original post

6 REPLIES 6

Lynda1
Kilo Sage

We migrated from another provider to ServiceNow Virtual Agent (VA) almost a year ago. The ServiceNow VA is as smart as you build it.

It can provide direct answers to a question if the intent score meets the threshold. You can build a fallback topic with questions and answers and have the "components" use the NLU so if the answer is found in the intents, the topic with the answer is displayed.

Not sure if you found the documentation about VA so providing a link here.

Conversational Interfaces Home (servicenow.com)

Since we have used ServiceNow for many years, their Virtual Agent was easier to implement and obtain information that relate to the users account. The conversation portion is fantastic.

Chris D
Kilo Sage
Kilo Sage

This I think is arguably one of the biggest weaknesses of ServiceNow's VA compared to others (to be honest, I'm thinking of one in particular I've seen but don't recall the name).

You can certainly build it into VA, but it'll require a bit of design and implementation effort.

As you mentioned, the fallback topic would be the place to start, but something to consider is if you're using AI Search, the AI Search Fallback does come first and that may or may not be helpful for you. In my opinion, using your exact use case, the ootb AI Search Fallback may not be too great because KB results don't show the content - just the Short Description and a real brief blurb with a link to the full article. Genius Results in theory could be useful to show (most) of the (short) article's content, but I still find Genius Results too unreliable and unpredictable.

There is a nice FAQ Conversation Builder topic block which basically does what you want (displays the entirety of the content in-line - not a hyperlink - from FAQ KB articles), but it is limited to FAQ articles specifically and you'd have to plug it into a custom (AI Search) Fallback topic using the user's utterance (vaSystem.searchText iirc) as the search input.

Thanks for the answer!

I'm peeking around my PDI but cannot find the FAQ KB type, something to be installed or create myself perhaps?

 

/Mikael

I'll go over it in my Academy tomorrow at 8:30am PT. You need to install the advanced Knowledge Management plugin to enable that article template type.