The CreatorCon Call for Content is officially open! Get started here.

Now Assist in Virtual Agent Direct Answer for Single Matching Topic

Jan Strama1
Tera Guru

Hi Everyone,

 

we have setup Now Assist in Virtual Agent and one of our Managers has a request to reduce the "Thinking Information" the Agent shows. They find it a bit overboard when there is a direct topic match for the request (see example below).

Do you know if there is a way to finetune/remove the details on the thinking steps?

JanStrama1_0-1761212962467.png

 

2 REPLIES 2

GlideFather
Tera Patron

Hi @Jan Strama1

 

I dont' have any available instance with Now Assist capabilities at the moment (only an old screenshot as seen below) so I cannot give you exact navigation but I will try to give you approximate one ;))

 

If I remember it correctly, you need to know which AI Agent does this... go to AI Agent Stuido > select you agent and then go to "Add tools an information" tab and there are several tools - e,g, flow action, script, record search.....

 

GlideFather_0-1761229094517.png

and if you open these tools one by one - such as Get Similar Records - there is the option "Display output" and I have a feeling that if this is marked to true then it is kind of explaining what it is doing... and perhaps this would be the way how to hide it. If not this, perhaps some other, try to review it, I don't remember it and cannot access it myself, but it could give you an idea.

 

You can also navigate to [sn_aia_tool] table to see all the tools and find some patterns there.

 

Let me know if this helped and tell me about your progress, good luck!

———
/* If my response wasn’t a total disaster ↙️ drop a Kudos or Accept as Solution ↘️ Cheers! */


rpriyadarshy
Giga Guru

Just to Add 

 

--> Have a look at the PROMPT which is used in the AI Agent against which Tools are running. Do Prompt needs fine tuning?

 

---> In Tools you see 

 

rpriyadarshy_0-1761281278646.png

 

 

Output transformation strategy
This option controls the verbosity and format of an agent tool's output when displayed to the end user. It is set within the configuration of a specific tool in the AI Agent Studio. 

 

  • None: Provides the raw, unprocessed output from the tool, which is often JSON or other technical jargon. This is most useful for system-level processing or when outputting a specific record ID.
  • Concise: Delivers a shorter, more direct output. This is ideal for giving quick confirmations to the user.
  • Paraphrase: Rewords the raw output in a more natural, human-readable way. This can be used to convert a technical message into a clear status update for the user.

 

  • Verbose: Generates a more detailed and eloquent answer, including additional context or explanation. This is helpful for user education or for providing more comprehensive information. 

 

Also Performance is Optimized if you update the Patch Level to  YP8 or  ZP2. Have a look at this video on below URL..

 

Fully agentic Virtual Agent in the October release... - ServiceNow Community

 

 

 

 

Regards

RP