Now Assist for CSM in Virtual Agent

FernandoCastro · ‎07-31-2025

Now Assist Virtual Agent for CSM transforms how customers interact with your support organization by enabling natural, multi-turn conversations that resolve issues faster—without requiring an agent. Built on large language models (LLMs), it understands user intent, retrieves relevant knowledge, and guides customers through resolution steps or service requests in real time.

Unlike traditional topic-based chatbots, this experience uses generative AI to dynamically generate responses, summarize complex topics, and power conversational flows such as case creation, catalog ordering, or guided troubleshooting. It integrates seamlessly into Service Portal, Engagement Messenger, and third-party messaging channels.

Designed specifically for customer-facing use cases, the Now Assist Virtual Agent in CSM helps deflect Tier 1 volume, shortens time-to-resolution, and delivers a more consistent experience while remaining easy to configure and govern with built-in tools like guardrails and analytics.

Key Capabilities:

Enhanced Chat: a dynamic modern window that allows multiple active conversation, modes and superior search capabilities by redirecting you from the search bar

Conversational Q&A: Summarize and synthesize answers to user questions using trusted knowledge sources.
Conversational Catalog: Allow users to request items using natural language (e.g., “Order a new laptop”).
Small talk and mid-topic switching: Engage in multi-turn conversations that feel natural and dynamic.
Sensitive topic fallback: Automatically redirect users to live agents when sensitive topics are detected.
Chat Streaming responses: real time stream instead of waiting for the entire message to render
Now Assist Topics: uses this description to find the best topic match for the user utterance.

Subflows and Actions: Create Now Assist Topics from Workflow Studio's subflow and actions that are conversational and trigger atomic automations
Enable Agentic Workflows (Group of AI Agents with a common mission) in Now Assist Virtual Agent

Implementation:

1. Get Started Here

2. Configure Now Assist for Customer Service Management (CSM) in Virtual Agent

3. Use Now Assist for Customer Service Management (CSM) in Virtual Agent for Scheduling

Key Best Practices

Formal Learning

Frequently Asked Questions (FAQs)

1. What is Now Assist Virtual Agent and how does it differ from traditional VA?

Now Assist Virtual Agent (NAVA) uses large language models (LLMs) to interpret customer intent, generate dynamic responses, and engage in multi-turn conversations without the need for intent training or rigid topic scripts. It’s more adaptive than traditional NLU-based VA and easier to configure.

2. What kinds of CSM use cases can NAVA support?

NAVA can support:

Knowledge Q&A using semantic and generative search
Case creation flows with smart in-form suggestions
Catalog ordering via conversational input
Agentic workflows, such as auto-triaging an issue or executing multi-step tasks
Escalations to live agents, with transcript handoff

3. How does NAVA interact with Agentic AI in CSM?

Agentic AI enables NAVA to do more than just answer questions. It can trigger autonomous workflows behind the scenes using AI agents. For example:

Auto-triage a case based on issue type
Run a document verification flow
Extract data entities from a message
Trigger a subflow to create or update records
These are orchestrated through Agent Orchestrators that plan and delegate tasks to worker agents, all initiated conversationally from NAVA.

4. Do I need to script or configure these agentic workflows manually?

Not fully. You can use Now Assist Skill Kit to configure the generative skill that triggers the workflow. Then define conditions, triggers, and flows via Flow Designer or Subflows. Some use cases like “Triage a case” or “Verify document” are available out of the box.

5. What if a customer query spans multiple topics or needs clarification?

NAVA supports contextual memory and multi-turn interactions, allowing it to ask clarifying questions, gather more information, or guide users through a dynamic path unlike static forms or deterministic VA topics.

6. How are fallback and sensitive topics handled in NAVA?

Through Now Assist Guardian, NAVA can detect when a customer raises a sensitive topic (e.g., harassment, personal information, health concerns). It can then:

Log or block the message
Redirect to a fallback topic or live agent
Apply language moderation or masking

7. Can NAVA trigger multiple skills or actions in a single interaction?

Yes. When paired with Agentic AI, NAVA can trigger a chain of actions depending on the customer’s intent. For example, “I need to change my billing address” can validate identity, update the record, and confirm changes all without human intervention.

8. How do I know when to use an agentic workflow vs. a traditional one in NAVA?

Use agentic workflows when the customer request requires reasoning, dynamic paths, or multi-step orchestration. Use traditional flows when tasks are deterministic, regulated, or require strict inputs. A hybrid model is common.

Scenario	Recommended Approach
“Reset my password”	Traditional catalog flow
“I received a wrong item, what do I do?”	Agentic triage and Q&A
“I need to update my payment info”	Agentic workflow with context checks
“I want to order a new laptop”	Conversational catalog topic

9. What configuration is required to get started?

You’ll need to:

Activate Now Assist for CSM and Virtual Agent plugins
Enable AI Search, Engagement Messenger, and Now Assist Panel (optional)
Configure catalog item support for conversational ordering
Review and update knowledge base filters and user criteria
Define fallback handling and guardrails in Now Assist Guardian

10. How is NAVA monitored and governed?

Use Now Assist Analytics and Guardian to:

Track session volume, deflections, and handoffs
Monitor sentiment and topic engagement
View acceptance and rejection trends by skill
Enforce guardrails for language, PII, or unsupported intents