Now Assist Response Length and Resolution Feedback

[Vinicius Andrad]

Hi everyone,

We are currently implementing Now Assist in Virtual Agent and received two questions/concerns from the customer regarding the user experience.

1. Synthesized responses are too long

The customer feels that the synthesized responses are often too detailed and lengthy, even though the Assistant is already configured with:

Tone: Direct
Response length: Succinct

For example, for a simple question such as *“What is churn?”*, the expectation would be a short answer with only the essential information, followed by an opportunity for the user to ask for more details.

Ideally, the experience would be something like:

User: What is churn?
Now Assist: Churn is the rate of customer cancellations or losses and is commonly used to measure customer retention. Would you like to know how it is calculated or what types of churn are tracked?

Instead, the current synthesized response may include several sections, detailed explanations, related topics, and multiple references.

Is there any additional configuration, prompt, property, or supported approach to further limit the length/detail of synthesized responses?

The goal is to have a more conversational experience: provide a simple answer first and progressively add more information only if the user asks follow-up questions.

2. Asking if the answer resolved the user's question / closing the conversation

The second concern is related to Enhanced Chat and Premium Chat.

The customer would like the Assistant, after providing an answer, to ask something like:

“Did this resolve your question?”

* Yes
* No

If the user selects Yes, the conversation could then be considered completed/closed and, ideally, trigger the configured post-chat survey.

This is similar to the experience they previously had with Standard Chat, where the conversation had a clearer closure flow.

With Enhanced/Premium Chat, we have not found an OOB way to consistently ask whether the response resolved the user's need after a synthesized answer and then explicitly close the conversation.

Is there a recommended OOB configuration for this behavior in Enhanced or Premium Chat?

We are aware of the thumbs up/down response feedback, but the customer's requirement is specifically to have an explicit “Did this resolve your question?” step that can lead to conversation closure and the post-chat survey.

Any guidance or recommended approach for these two scenarios would be appreciated.

Thanks!

[KPNow]

To tame those lengthy synthesized responses, relying solely on system-level settings can often fall short because the underlying LLM still tries to be helpful by over-explaining. The best way IMO to enforce a truly conversational, multi-step flow is to customize the specific System Prompt inside Now Assist Admin > Generative AI Controller > Prompts (or within the prompt instructions for Contextual Search / KB Q&A). In your system prompt instructions, explicitly define strict constraints such as: "Limit your initial answer to two sentences maximum. Focus strictly on a high-level definition and end with a concise follow-up prompt offering to explain further." Furthermore, if you are pulling content from Knowledge Base articles, ensure your KB chunking and response generation settings prioritize summary snippets rather than attempting to synthesize full-length articles.

For the explicit "Did this resolve your question?" closure loop in Enhanced/Premium Chat, standard OOB behavior defaults to the lightweight thumbs up/down feedback widget, but you can build a more definitive workflow using the Virtual Agent Designer. In your Virtual Agent topic (or fallback topic), route the output of the Now Assist node directly into a standard User Input (Boolean / Static Choice) control that asks the user if their issue was resolved. If the user selects "Yes," route the flow to an End Conversation topic node, which naturally triggers your configured Post-Chat Survey in the messaging interface. If they select "No," you can seamlessly branch the logic to present a follow-up prompt, search another source, or trigger a live agent transfer via Advanced Work Assignment (AWA).