From Chat to Calls: Introducing Agentforce Voice

From Chat to Calls: Introducing Agentforce Voice

By

Your customers talk to AI every day. They ask their phones for directions, tell their smart speakers to set timers for cooking, and expect voice interfaces to be available for questions. But when they call your contact center, they end up in a queue.

As a Salesforce Admin, you’re the one who decides how your customers experience your company. You’ve built the flows, configured the channels, and wired up the automations that make service feel seamless. Agentforce Voice is the next thing to add to that list.

It brings the AI agent you’ve already been building — the one that understands natural language, looks up records, and resolves issues — to the phone channel. It uses the same concepts you know in your builder, with subagents and actions, but adds a new channel your customers already use every day.

Let’s break down what it is, what you need to set it up, and how to make sure your first rollout is smooth.

Meet Agentforce Voice

Agentforce Voice is a channel, not a separate product. It’s the same agent-building model you already know: subagents, actions, and Agentforce Builder applied to inbound phone calls instead of chat or messaging.

The infrastructure underneath it is Salesforce Voice (formerly Service Cloud Voice). That layer manages the telephony connection: your Amazon Connect instance or partner provider, audio streaming, and speech-to-text transcription. Agentforce Voice sits on top of that and handles the reasoning: what the caller is asking, what to do about it, and when to bring in a human.

Here’s what a typical call workflow looks like once Agentforce Voice is up and running.

  1. A customer calls your contact center number.
  2. Agentforce Voice picks up and greets them, in a voice you select and can preview before it goes live.
  3. The agent understands natural speech.
  4. The agent resolves the issue using your subagents, actions, and Flow (or Apex) backed logic.
  5. If the caller wants to speak to a human, or the agent reaches its limit, it transfers the live call with the full transcript and a prepopulated case already attached to the representative’s.

Understand the foundation

Three layers work together, and each has a setup role.

Agentforce Voice stack

  • Your telephony provider — Amazon Connect or any of the 20+ supported Contact Center as a Service (CCaaS) partners — handles the phone numbers, audio streams, and call routing at the carrier level.
  • Salesforce Voice connects that provider to Salesforce: audio streaming, real-time speech-to-text transcription, and the live call object the Agentforce layer works from. This has to be stable before you add an AI layer on top.
  • Agentforce Voice sits above that. It receives the transcribed speech, runs it against your subagents and actions, and responds in the voice you selected and previewed before go-live. This is what you build and configure in Agentforce Builder.

Give swift answers with automation

The calls that are best suited for Agentforce Voice share a simple profile: The answer already lives in Salesforce, the resolution is clean, and the customer mostly needs speed.

Solid examples for automation include:

  • Account balance checks
  • Reservation cancellations
  • Shipment tracking
  • Store hours and locations 

What these use cases have in common is that they don’t need a skilled rep to provide a correct answer. The only thing that’s required is access to data already in your org. Agentforce Voice can handle them in under 2 minutes, around the clock, without a rep having to work out of their queue.

For calls that do need to be transferred to a human, the agent provides an improved experience too. The rep who picks up an agent-escalated call sees the full transcript, a summary, and whatever case was created during the conversation. No need to ask a customer to recap. The rep can immediately pick up the conversation and start to solve the problem.

You don’t have to automate everything at once. Start with the two or three call types where resolution is fastest and most predictable. Validate those within the tools Builder gives you, and expand from there. That’s how the orgs that have done this successfully approached it and it’s best practice for any type of change, especially when the channel is as public as a voice agent. 

Get started with Agentforce Voice 

An important note before we get into setup: Agentforce Voice requires Salesforce Voice to be working underneath it. This isn’t a feature you can layer on top of a legacy telephony setup. If your org is still on a standalone Avaya or Genesys implementation with no Salesforce Voice integration, that’s the first conversation to have.

If you’re on Salesforce Voice, or you’re ready to get there, here’s what you need.

  • Salesforce Voice enabled and telephony provider configured (Salesforce Voice Setup)
  • Supported provider: Amazon Connect (native, Salesforce manages provisioning), or a CCaaS partner via Partner Telephony; CCaaS refers to cloud-hosted contact center platforms like Genesys, NICE, Five9, and Vonage that integrate with Salesforce Voice to handle telephony. (Bring Your Own Channel for CCaaS)
  • Agentforce for Service license
  • Omni-Channel enabled
  • Admin permissions: Customize Application, Manage Call Centers, Contact Center Admin

For the quick setup steps (creating your contact center, adding the voice channel, configuring subagents and escalation), complete the Agentforce Voice badge.

Explore what’s coming

Agentforce Voice is generally available now, but the voice channel is expanding.

  • Voice for Digital Channels (GA Q2 2026) brings the same capability to web chat, mobile apps, WhatsApp, and messaging channels. A customer can start typing in a chat window and switch to speaking mid-conversation without losing context. One agent, one configuration, working across every surface. The “build once, deploy anywhere” promise extends to voice too.
  • Agent Script for Voice (Q1 2026) adds structured, scripted conversation flows as an option alongside fully autonomous AI responses. If you have use cases where you need predictable, word-for-word phrasing (regulated industries, compliance-sensitive disclosures, specific call types where variance isn’t acceptable), Agent Script gives you that guardrail without abandoning the channel.
  • Voice in the Salesforce mobile app (Pilot) helps internal users (reps in the field, account executives between meetings) access CRM data, summarize accounts, and trigger workflows by speaking directly into the Salesforce app. Same agent, different direction: customer-facing vs. employee-facing.

If you’re building for voice now, these aren’t distant features to plan around. They’re the next steps in the same setup you’re already learning.

Resources

How Agentblazers Are Solving Real-World Problems with Agentforce

How Agentblazers Are Solving Real-World Problems With Agentforce

At TDX 2026, the Agentforce Hackathon brought together builders from across the Salesforce ecosystem to create AI-powered solutions that tackle real-world challenges. And the ideas were anything but theoretical. This year’s finalists explored how Agentforce can: Help care coordinators better support patients. Give city teams earlier visibility into infrastructure issues. Make knowledge experiences more dynamic […]

READ MORE
Introduction to Salesforce Headless 360 for Admins

Introduction to Salesforce Headless 360 for Admins

Salesforce Headless 360 is starting to show up in conversations with your developers and leadership. If you’ve been wondering what it actually means for your day-to-day life, or whether it changes anything you do, you’re in the right place. The short answer: It makes the work you already do more powerful and more reusable. Your […]

READ MORE
Setup with Agentforce Is Now Generally Available

Setup with Agentforce Is Now Generally Available: Built With Admin Feedback at the Center

At TDX 2026, one thing became incredibly clear: The response to Setup with Agentforce was overwhelmingly positive. Many admins already reported great success trying out the Setup with Agentforce beta for tasks like creating objects, troubleshooting user access, and getting information on Salesforce features.  At the same time, across demos, customer conversations, and True to […]

READ MORE