What Is Agentforce Voice? Salesforce’s AI Voice Agent Explained (2026)

Your contact center handles thousands of calls a day. Most of them are routine: order status checks, password resets, appointment reschedules. And yet, each one costs your business $5 or more to resolve with a human agent. That math stops making sense very quickly.

Enter Agentforce Voice. Salesforce’s answer to the oldest problem in contact center automation. Launched in October 2025, it brings Salesforce’s AI voice agent capabilities directly into your CRM, telephony stack, and customer workflows, replacing the rigid, menu-driven IVR systems that have frustrated customers and cost center leaders for decades.

This guide covers everything: what it actually is, how it technically works, where it genuinely shines, what the real pricing looks like, and the limitations competitors will not tell you about.

What Is Agentforce Voice?

Agentforce Voice is Salesforce’s AI-powered phone agent. It picks up inbound calls, listens to what customers say, takes action inside Salesforce, and hands off to a human agent when the situation calls for it.

The easiest way to think about it: if you already use Agentforce to handle chat or messaging, Agentforce Voice brings that same intelligence to the phone. The same customer data, the same workflows, the same AI. You set it up once, and it works across your phone line, your website’s click-to-call button, and your mobile app.

The big difference from a traditional phone menu system is that customers do not have to press 1 for billing or press 2 for support. They just talk. They can explain their issue in their own words, change direction mid-call, or ask follow-up questions, and the agent understands them because it is pulling context from their actual account in real time.

What Makes it Different from a Chatbot?

A chatbot reacts. Agentforce Voice reasons. Using the Flash Planner, Salesforce’s purpose-built reasoning engine for voice, the agent does not just pattern-match keywords. It understands intent, retrieves relevant data from Data Cloud and CRM records, executes multi-step workflows (like logging a case, updating an order, or scheduling a technician), and decides in real time whether to resolve the issue or escalate to a human, along with full context and sentiment data.

“We don’t want the user to feel like they’re just talking to a robot. We want a system that recognizes not just what the caller says but the meaning behind how they say it.” John Emmons, Director of Research, Salesforce

Also Read: All About Agentforce: Features, Benefits, and Ksolves Expertise

How Does Agentforce Voice Work?

When a customer calls in, four things happen in sequence.

Step 1: It Listens and Converts Speech to Text

As soon as the customer starts speaking, Deepgram’s technology converts their words into text in real time, fast enough that the agent can start processing the question before the customer has finished asking it.

Step 2: It Figures Out What the Customer Wants

This is where the Flash Planner comes in. It reads the text, works out what the customer actually needs, pulls up their account information from Salesforce, and decides what to do next, all in a matter of milliseconds. This is what makes the conversation feel like a real back-and-forth rather than a slow, robotic exchange.

Step 3: It Speaks the Response Back

The agent’s reply is converted into natural-sounding speech using ElevenLabs. You can customise the voice to match your brand’s tone, so it does not sound like a generic automated system.

Step 4: It Knows When to Involve a Human

If a customer sounds frustrated, keeps repeating themselves, or has a problem the agent cannot solve, the call gets passed to a human agent. That person receives the full conversation history, a sentiment note, and the customer’s account details before they even say hello. The customer never has to start over.

Agentforce Voice vs. Traditional IVR

Most businesses running legacy IVR systems (Nuance, Genesys, Avaya) are automating 20 to 30 percent of their call volume at best, and doing so with experiences customers actively hate. Here is how Agentforce Voice compares across the metrics that matter.

Also Read: Salesforce Agentforce vs. Traditional Support: Is It A Game Changer

Use Cases of Agentforce Voice by Department

Agentforce Voice is most commonly positioned as a customer service tool, but its use cases extend across the business. Here is where it creates the most tangible value.

Customer service and support

• Order status and tracking queries
• Returns and refund processing
• Password reset and account updates
• Appointment scheduling
• Tier-1 and Tier-2 issue resolution
• Technical troubleshooting using knowledge base articles

Sales and lead qualification

• Inbound lead qualification via voice
• Meeting scheduling with sales reps
• CRM record creation and enrichment
• Product inquiry routing
• Outbound sales calls are in development and not yet generally available

Field service and operations

• Technician dispatch scheduling
• Service window confirmations
• Outage notifications for utilities
• Delivery status and exception handling
• Real-time inventory queries

Internal and employee-facing

• Real-time sales coaching during live calls
• Knowledge base retrieval for human agents
• HR query handling (leave, payroll questions)
• Onboarding Q&A automation
• Compliance script adherence checks

Agentforce Voice in Action: A Real-World Example

Ksolves deployed Agentforce Voice for an unattended laundry service provider, a business running self-service facilities with no staff on site. When machines broke down, coins got stuck, or receipts went missing, customers had no one to turn to.

Ksolves built a 24×7 voice concierge on Agentforce for Voice, connected to Salesforce Data Cloud and live IoT data from the machines. Customers simply call in and speak naturally. “I need my receipt.” “Machine 5 is leaking.”

The agent understands, checks the customer’s account and the live machine status, and acts instantly. Receipts are sent automatically, refunds processed in real time, and troubleshooting steps delivered verbally. If a customer uses a distress keyword, the call escalates to a live human agent immediately, with full context already loaded.

The results:

• 85% of refund and receipt requests were resolved by the voice agent with zero human involvement
• 30% reduction in operational costs from fewer site visits and lower support volume
• Instant refunds during the call, replacing a slow manual process
• Faster safety response through automatic emergency escalation

Read the full case study. 

Set Up Requirements and Implementation Timeline

Setting up Agentforce Voice is a two-layer process: configuring the AI agent itself, and connecting it to your telephony infrastructure via an Omni-Channel flow.

Step 1: Prerequisites

You need Salesforce Enterprise or above, a Service Cloud license, Salesforce Voice enabled, a CTI telephony partner (Amazon Connect, Genesys, Five9, NICE, or Vonage), and Agentforce enabled in your org. Data Cloud is technically optional but strongly recommended for real-time personalization.

Step 2: Create your Voice-enabled Agentforce Agent

In Salesforce Setup, navigate to Agentforce Agents. Use the Service Agent template or build a custom one. Define topics (the categories of questions your agent handles) and actions (the specific tasks it can execute). Enable voice under Connections and select Telephony as the connection type.

Step 3: Configure your Omni-Channel Inbound Flow

Build a flow that routes incoming PSTN calls to your Agentforce Voice agent. This flow also defines the escalation logic: when and how calls transfer to a human agent, and what context (transcript, sentiment, CRM data) passes with them.

Step 4: Test, Iterate, and Go Live

Use the preview feature in Agentforce Builder to simulate voice interactions before going live. Monitor average handle time, first call resolution, and call deflection rate closely in the first 30 days. The Flash Planner improves with each interaction as your data volume grows.

Security, Data Privacy, and Zero-Retention Policy

For compliance officers and regulated industries, this section matters as much as features. Agentforce Voice is built on Salesforce’s Einstein Trust Layer, with specific data handling agreements for its third-party AI providers.

Salesforce has established zero data retention policies with all three of its AI voice partners.

Deepgram handles speech-to-text transcription. Under the zero-retention agreement, your audio and transcripts are not stored or used to train their models.

OpenAI is listed as a zero-retention partner within the Agentforce Voice pipeline. Your data is not stored or used to train their models. 

ElevenLabs handles text-to-speech voice synthesis. Your content is not retained or used for model training.

This means customer voice data flows through these providers in real time, but is not stored outside Salesforce. Transcripts and call summaries are stored within your Salesforce org, subject to your existing data governance, field-level security, and retention policies.

Looking to Deploy Agentforce Voice? Ksolves Can Help

Agentforce Voice is a powerful product, but getting it right takes more than just switching it on. The quality of your Salesforce data, the design of your escalation flows, your choice of telephony partner, and how well your agent topics are defined all directly affect what results you see.

Ksolves is an Agentforce Voice Implementation partner with hands-on experience implementing Agentforce Voice across industries. From initial scoping and architecture to agent configuration, telephony integration, and post-launch optimization, our team handles the full deployment so your business can focus on the outcomes rather than the setup.

Whether you are evaluating Agentforce Voice for the first time, migrating from a legacy IVR system, or looking to expand an existing Agentforce deployment to the phone channel, we can help you get there faster and with fewer surprises.

Get in touch with the Ksolves Salesforce team to discuss your requirements.

FAQs

Q1. Do I need Salesforce to use Agentforce Voice?
It is a native Salesforce product and requires a minimum of a Salesforce Enterprise license with Service Cloud and Salesforce Voice enabled. Not on Salesforce yet? Talk to Ksolves to understand what the setup would involve.

Q2. How long does it take to set up?
Simple deployments take four to six weeks. Complex ones with custom flows and Data Cloud integration can run eight to twelve weeks. Working with a partner like Ksolves from the start keeps timelines on track.

Q3. What languages does Agentforce Voice support?
English only, as of March 2026. Multilingual support is on Salesforce’s roadmap, but has no confirmed date.

Q4. Is customer data safe with Agentforce Voice?
Yes. Salesforce has zero data retention agreements with Deepgram, OpenAI, and ElevenLabs. Your data is processed in real time but never stored outside your Salesforce org.

Q5. How do I know if it is the right fit for my business?
If you are on Salesforce and handling high volumes of routine inbound calls, it is worth a serious look. Ksolves offers Agentforce consulting to help you assess readiness before you commit. Reach out at sales@ksolves.com.