Meet Agentforce Voice
Learning Objectives
After completing this unit, you'll be able to:
- Describe the key functions of Agentforce Voice.
- Outline the main steps to create a voice-enabled agent.
- Understand the built-in guardrails and trust policies.
Discover Agentforce Voice
Imagine a customer support team that is available 24 hours per day, can understand your customers’ needs and instantly solve their problems, and never needs to sleep, eat, or take breaks. That’s the magic of Agentforce Voice. It’s a conversational AI feature that gives your Agentforce service agents the ability to speak with customers, understand what they’re saying, and take action.
Here’s how it works: A customer calls in with a question and Agentforce Voice immediately gets to work. First, it uses a speech-to-text model to turn the customer’s spoken words into text. Then a latency-optimized reasoning engine, called the Flash Planner, figures out what the customer wants, and crafts a text response. Finally, a text-to-speech model converts that text back into natural-sounding audio for the customer to hear. All of this happens in a rapid succession, so the conversation can take place in a natural, free-flowing way.
Create a Voice-Enabled Agent
Let’s take the example of using your agent to handle telephone calls from your customers. You start with a new agent template, add a Telephony connection, and then the fun part begins: choosing a voice! You can select a voice based on gender, age, or accent, and even fine-tune its speed and stability (how consistent the voice is between each generation) to match your brand's personality.
Once you've customized its topics and added a welcome message, you can preview and test how the agent sounds before deploying it.
Setting up the call flow is just as important. You need to configure an inbound call transfer using an Omni-Channel flow. This tells your phone system to send incoming calls to the correct voice-enabled agent. And what happens if the AI agent can’t solve a customer's problem? That's where call escalation comes in. You can set up another flow that allows the agent to gracefully hand off the conversation to a human representative. This ensures your customers always get the help they need, whether from an AI or a person.
Agentforce Voice and Service Cloud Voice work together to route your customer calls to an Agentforce agent. Service Cloud Voice is the bridge that connects your third-party phone systems to Salesforce. Agentforce Voice is the AI brain that handles the actual conversation. Together, they transform your standard call center into an intelligent, round-the-clock support hub.
Converse Safely with Guardrails
But it's not just about talking; it’s about talking well. Agentforce Voice comes with built-in guardrails and instructions to make sure its responses are safe, clear, and helpful. This includes:
- Optimizing spoken delivery: The agent knows to pause for clarity, like when reading out an address.
- Providing concise responses: Word limits keep the agent from rambling.
- Ensuring content safety: Guardrails help prevent offensive language and protect your company's private information.
Keep Your Data Safe with Zero Data Retention
Salesforce has established zero data retention policies with its partners, including OpenAI, Deepgram, and ElevenLabs. This means your data isn’t stored by these third-party providers or used to train their models, keeping your information secure.