The Get Case Status and Manage Cases AI Voice Agent is an out-of-the-box capability in Now Assist for CSM that enables customers to check the status of an existing case and create new cases through natural voice conversation — without waiting on hold or speaking to a live agent.
The voice agent integrates with your CCaaS telephony provider, authenticates the caller, retrieves live case data directly from the Now Platform, and responds conversationally — handling follow-up questions, repeating information on request, and escalating to a live agent with full context when needed.
Overview video:
Key Capabilities
-
Secure caller authentication: Verifies caller identity using PIN (SoftPin), SMS code, Knowledge Factors, OTP, or Okta Push Notification before accessing any case data.
-
Case status retrieval: Retrieves live case state, priority, assigned group, product, and latest work notes from CSM and reads them back conversationally to the caller.
-
Multi-turn conversation: Supports natural follow-up questions within the same call — callers can ask for specific fields, request a repeat, or pivot to a new question without restarting the interaction.
-
New case creation: Collects issue description from the caller and creates a new CSM case in real time, confirming the case number before ending the call.
-
Graceful escalation: Transfers the caller to a live agent when needed, passing the full conversation transcript, summary, and detected intent directly into the agent workspace.
-
Post-call analytics: Logs full transcripts to Interaction records and tracks deflection rates, conversation outcomes, and post-call CSAT scores (1–5 scale) in the Voice Analytics dashboard within Assistant Designer.
OOTB CSM Voice Agents included (as of Q1 2026):
- Case Status Voice Agent
- Create Case Voice Agent
Implementation
1. Prerequisites
Before configuring the AI Voice Agent, ensure the following are in place:
- Now Assist for CSM is activated on your instance (Pro+ or Enterprise+). The platform-level AI Voice Agent plugin is auto-installed with Now Assist for CSM — no separate installation is required.
- A supported CCaaS integration is configured. As of Q1 '26, Genesys Cloud, 3CLogic and Twilio are the supported providers. Amazon Connect, Five9, and NICE are planned for H2'26.
- Instance is on Zurich Patch 4 or Yokohama Patch 11 minimum.
1.1 Configure the Voice Assistant in Assistant Designer
The Voice Assistant is the entry point that connects your telephony channel to the Now Platform and defines authentication and personality settings.
- Navigate to Now Assist > Assistant Designer and create a new Voice Assistant.
- Select your CCaaS provider (Genesys Cloud or Twilio) and enter the integration credentials from your CCaaS account.
- Define your authentication method under Security settings:
- PIN (SoftPin) is recommended for most B2C deployments — the caller enters their PIN via phone keypad followed by the pound key, validated against the User table on the Now Platform.
- Alternative methods: SMS Code, Knowledge Factors, OTP (Auth App), Okta Push Notification.
- Set the voice assistant name, greeting language, and voice personality. Supported languages include English, German, Spanish, French, French Canadian, Brazilian Portuguese, Dutch, Italian, Korean, and Mandarin. ServiceNow uses the Cartesia Sonic-3 TTS model — ASR selection is managed by ServiceNow and is not customizable.
1.2 Review and activate the OOTB CSM Voice Agent in AI Agent Studio
The Case Status Voice Agent & Create Case Voice Agent is pre-configured as part of the CSM OOTB collection and covers both case status retrieval and new case creation.
- Navigate to AI Agent Studio and locate them under the CSM Voice Agent collection.
- Review the agent's description, role instructions, and pre-configured tools. The OOTB agent uses record operation tools to read case data from the Case table — no custom tool configuration is required to get started.
- Optionally extend the agent by adding:
- RAG Search tool — to surface and summarize relevant knowledge articles during the call. If adding KB search, filter to the specific knowledge bases relevant to your voice channel use cases rather than an open-ended search across all KBs to maintain response latency.
- Subflows or script tools — to trigger additional workflows on case creation or update.
- MCP Server tool — to connect to external agentic systems if needed.
1.3 Map Voice Agent to Voice Assistant
Once the agent is reviewed and configured:
- Return to Assistant Designer and add the CSM Voice Agent to your Voice Assistant configuration.
- The Voice Orchestrator routes caller utterances to the LLM, which determines the next response and which tool to invoke based on the caller's intent and the agent's instructions — no predefined dialogue trees required.
- Configure fallback behavior for scenarios where the AI cannot resolve the request or a call exceeds the 10-minute limit:
- Transfer to live agent — recommended for most deployments. The caller is transferred with a transcript summary and detected intent pre-populated in the agent workspace.
- Generate ticket via Record Producer — an alternative fallback that creates a case automatically if no live agents are available.
1.4 Test the Voice Agent
Use a live call to the configured phone number to validate the full conversation flow. A dedicated in-platform voice testing experience is on the roadmap for Q2 '26. In the interim, review call outcomes using:
- Interactions table — full transcript logged per call as soon as the call ends (caller and agent turns).
- Execution Plans table — step-by-step tool invocation log for debugging unexpected agent behavior.
- Analytics dashboard — accessible from the Analytics tab in Assistant Designer. Tracks total voice conversations, deflected conversations, conversation outcomes, satisfaction scores, and performance of AI agents invoked by voice assistants.
Recommended edge cases to validate before go-live:
- Caller provides an invalid case number → agent should inform the caller it cannot locate the case and prompt for a correct number.
- Caller provides a case number belonging to another account → in the initial release, case ownership validation is not enforced; communicate this limitation to your implementation team. Case ownership validation is planned for a future release.
- Call exceeds 10-minute fallback threshold → confirm fallback behavior triggers correctly.
Key Best Practices
- Keep case data current and complete in CSM — the voice agent reads directly from live records, so data quality directly affects response accuracy.
- For high-volume deployments, consider routing the AI Voice Agent as a dedicated option within your existing CCaaS IVR tree (e.g., "Press 2 for case status") until deflection rates are validated, rather than replacing the full IVR immediately.
- Monitor the Analytics dashboard weekly in early production — review deflection rate, conversation outcomes, and CSAT to identify gaps in agent instructions or tool coverage.
- Use PIN (SoftPin) as the primary authentication method for phone-based B2C interactions — it is the most frictionless option and is validated natively against the User table.
- Scope the Search Retrieval tool to 2–3 knowledge bases most relevant to your voice channel use cases — avoid open-ended search across all KBs to maintain low response latency.
- Review Execution Plans regularly during the first weeks of production to understand step-by-step tool invocation patterns and tune agent instructions accordingly.
Formal Learning
- AI Agents Delivery Accreditation
- CRM AI Academy: Voice Agents - From Intent to Real-time Resolution
- AI Voice Agents — Community Overview & FAQ
- Integrating Voice Service with CCaaS Providers — Product Documentation
Measured Success
| Outcome | Explanation (with applicable use case) | Success Metric |
|---|---|---|
|
Reduced inbound call volume |
Status inquiry calls handled by the AI Voice Agent without live agent involvement, freeing agents for complex issue resolution |
% of calls deflected by voice AI |
|
Faster case status delivery |
Callers receive accurate case status in under 60 seconds with no queue wait, 24/7 including after hours |
Average time to case status (seconds) |
|
Consistent customer experience |
Every caller receives the same structured, accurate response regardless of time of day or agent availability |
Post-call CSAT score (1–5 scale) |
|
Reduced average handle time |
Live agents spend less time on routine status lookups and focus on resolution work that requires human judgment |
AHT reduction (minutes) |
|
Improved after-hours coverage |
24/7 availability eliminates the after-hours gap where callers previously had no self-service option for case status |
% of after-hours calls resolved by AI |
|
New case creation accuracy |
Cases created via the voice agent are complete and correctly categorized, reducing downstream triage rework |
Field completion rate on voice-created cases (%) |
Frequently Asked Questions (FAQs)
1. Is this a custom-built agent or an out-of-the-box capability?
This is an out-of-the-box (OOTB) AI Voice Agent delivered as part of Now Assist for CSM. The
- Case Status Voice Agent
- Create Case Voice Agent
are pre-configured and ready to activate — no custom agent build is required to get started. You may extend it with additional tools or instructions via AI Agent Studio.
2. What happens if the caller provides an invalid case number?
The AI Voice Agent informs the caller it cannot locate a case matching that number and prompts them to provide the correct case number. The call does not terminate — the caller can retry or ask to speak to a live agent.
3. How does the conversation flow work? Is it scripted?
The conversation is not scripted. The Voice Orchestrator routes caller utterances to the LLM, which dynamically generates the next response based on the caller's input, the configured agent instructions, and the tools available. Callers can ask follow-up questions in any phrasing, ask for repetition, or change topics mid-call — all handled naturally without predefined dialogue trees.
4. How many Now Assist assists does each call consume?
As of the March '26 release (ZP7/AP1+), assists follow a tiered Agentic Workflow model: Small (<4 tool actions), Medium (5–8 actions), Large (9–20 actions). Calls under 30 seconds consume zero assists. For instances on versions prior to March '26, a flat 50 assists per call applies. Check the Now Assist Overview publications for current rates.
5. Can the voice agent sit within an existing CCaaS IVR tree rather than replacing it?
Yes. Two common configurations are supported: (1) keep the CCaaS provider's IVR tree and route a specific option (e.g., "Press 2 for case status") to the ServiceNow Voice Agent, or (2) remove the IVR tree entirely and route the dedicated line directly to Now Assist Voice. The right choice depends on your telephony architecture and call volume.
6. Are call transcripts stored on the ServiceNow instance?
Yes. Full transcripts (caller and agent turns) are logged as Interaction records in the Conversation table as soon as each call ends. Audio recordings are not stored on the instance — only text transcripts, which have negligible storage impact.
7. What model providers are supported?
Azure OpenAI, Google Gemini, and Anthropic Claude are supported. As of the March '26 release, NowLLM is also supported (English only). The ASR (speech-to-text) model is managed by ServiceNow and is not customizable by admins. TTS uses the Cartesia Sonic-3 model.
