ServiceNow Voice Agents: AI Agents communicating your solutions

[SujanDutta]

All you need to know about ServiceNow Voice Agents

In this series of episodes, we will have a complete walkthrough of Voice Agents, discussing the benefits, framework, set up guide and a quick preview, everything that you need to get started.

Episode 1: ServiceNow Voice Agents: Insights, Architecture & Demo

Your support team knows the problem. Calls are expensive. A single inbound call costs roughly 2.5 times more to handle than a chat or email interaction. At scale, that's brutal, especially when call volumes are high and wait times push customers toward frustration. ServiceNow Voice Agents solve this by letting AI handle the intake, qualification, and even resolution of common support calls. 

The Real Economics

When you have large volumes of self-service calls: password resets, ticket status checks, simple troubleshooting, human handling becomes a cost center drag. High wait times compound the problem. A frustrated customer waiting on hold is a customer you're about to lose.

Voice Agents don't eliminate human support. They redirect traffic. High-touch, complex issues still go to humans. But the volume of calls that can be resolved by an AI agent: quickly, accurately, and without a wait, changes your operational math.

How It Works: The Full Flow

The architecture is straightforward but elegant:

1. Telephony Gateway: A call comes in. ServiceNow connects to either Twilio (developer-friendly, API-first) or Genesis Cloud (plug-and-play for ops teams). No proprietary phone system required.
2. Speech-to-Text: The agent listens using a solid open model that handles accents and noise reasonably well and converts the caller's words into structured requests.
3. Orchestration & Tools: This is where the intelligence lives. The agent decomposes the ask ("my laptop won't start") into actionable steps. It has access to ServiceNow flows and custom tools you define. It can check ticket status, create incidents, run troubleshooting scripts, pull knowledge articles.
4. Action: The agent executes. Creates that incident, sets urgency, captures all the details.
5. Text-to-Speech: The agent speaks back to the customer with the outcome. Clear, natural, complete.
6. Logging: Every interaction gets recorded as an Interaction record in ServiceNow. Full transcript. Related tickets. Everything auditable and searchable.

What You'll need:

• ServiceNow Washington release or later.
• Telephony connector (Twilio or Genesis).
• AI capabilities configured, List of plugins(as per the required modules):
  • Now Assist for Platform (sn_genai_platform) for enabling default platform AI voice agents.
  • IT Service Management AI voice agent collection (sn_itsm_voice_aia) for enabling default ITSM AI voice agents. See Agentic AI in the Voice application for more information.
  • HR Voice AI Agents (sn_hr_voice_aia) for enabling default HRSD AI voice agents. See HR AI voice agents for more information.
  • CSM Voice AI agents (sn_cti_csm_cnt).
• AI search configured.

The key point: this isn't a locked-in black box. You design the agent's behavior. You choose what tools it has access to. You define the escalation path. It's as flexible as your ServiceNow configuration.

The Demo: What an Actual Call Looks Like

We watched an agent handle a real scenario. Caller: "My laptop won't start." The agent asked clarifying questions (lights? sounds?), diagnosed the issue in three exchanges, created a ticket with the right urgency, and closed the loop.

Total time: about 90 seconds. Total human involvement: none.

The interaction record captured everything: transcript, sentiment, related ticket, resolution status. The support team saw what happened, why the ticket was created, and where to pick up if escalation was needed.

That's not magic. That's design. The agent had a clear set of tools (create incident, check status, access KB articles) and a flow that walked through the logical steps.

Why This Matters Now

Three reasons:

1. Cost: Reducing call handle time and automating high-volume, low-complexity calls directly improves your cost per interaction and your SLA metrics.
2. Scale: You don't hire another 10 support agents to handle peak call volumes. You scale the agent.
3. Omnichannel Reality: Chat and email are commoditized. Call handling is still a pain point. Closing that gap on one platform, matters for retention, experience, and ops efficiency.

Episode 2: How to Configure ServiceNow Voice Agent: Setup Guide

If you've already built an agent flow in ServiceNow, configuring Voice Agents isn't a big leap. If you're new to it, the five-step process is straightforward enough that you can get it working without a lot of guesswork.

Here's what actually matters when you're setting it up.

Step 1: Basics Details

Start by creating an assistant under Conversational Interfaces > Assistants. Give it a name, add a description, and tag it in Basic details. The tags matter more than they seem, they'll help you track usage and performance in analytics later. 

Step 2: Voice Personality

ServiceNow supports a bunch of languages now, including Hindi. Pick the language your users actually speak. Then choose a persona, basically, whose voice do you want answering the phone? You can listen to the available personas and pick the one that doesn't sound like it's reading a script. The persona options change based on language, and they're all baked in.

Step 3: Communication Channels

Right now, Twilio and Genesys are available. If you're using Twilio, you need to connect the dots between ServiceNow and your Twilio account. Get a phone number from Twilio, grab the webhook URL from your assistant in ServiceNow, paste it into Twilio's voice configuration, copy back the authentication token, and paste that into ServiceNow. It sounds circular, but it's five minutes of copy-paste and you're done.

Step 4: Authentication

This is where a lot of people skip steps and regret it. You need two things: identification (is this person who they say they are?) and authentication (can they prove it?).

For identification, you can ask knowledge-based questions. ServiceNow asks something only the real user would know, they answer, and boom, you've got verification without a password prompt. For authentication, soft PINs work fine. It's a six-digit code the user sets up in their profile. When they call, the voice agent asks for it. 

Step 5: Safeguards

The agent won't always have an answer. Sometimes the conversation gets stuck or the user needs something the bot can't do. You've got two options: escalate to a live person immediately, or create a ticket and have someone call them back. Most setups use both depending on the scenario, urgent issues get live escalation, routine stuff becomes a ticket. Set your call duration limits and inactivity timeouts so calls don't just hang there indefinitely.

That's the configuration. It's not complex, and once you understand the flow, it's repeatable.