Build Your First AI Voice Agent in the Syllable Console

Getting started with AI agents does not need to be complex. In this walkthrough, Chris walks you through building a complete, working AI voice agent from scratch using the Syllable Console. By the end, you will have an agent that can answer customer questions over the phone, reference structured data, and transfer calls to a human, all without writing any code.

This tutorial uses a simple example: a support agent for a fictional camera rental business called Chris’s Splendid Shutter Shop. While the scenario is lightweight, the same workflow applies to production agents used in healthcare, customer support, sales, and operations.

Understanding the Syllable Console

When you log into the Syllable Console, you are presented with a workspace-driven interface. On the left-hand side, you will see your primary workspaces and resources.

Syllable agents are built using reusable components called resources. These resources define how an agent behaves, what it says, what it knows, how it sounds, and how it communicates with users.

The Core Building Blocks of an AI Agent

Every AI agent in Syllable is composed of a small set of foundational resources:

• Prompt: Instructions that define the agent’s role, behavior, and decision logic

• Message: The opening greeting spoken to the user

• Data Source: Structured or unstructured information the agent can reference to answer questions

• Voice Group: The language and voice used by the agent

• Channel: The communication method, such as a phone number

Each resource is created independently and then linked together when building the agent.

Step 1: Create the Prompt

The prompt is the most important part of your agent. It tells the agent who it is, what it should do, and how it should behave under different conditions.

Create a new prompt and give it a clear name, such as Chris’s Splendid Shutter Shop Agent. Choose your model provider, model, and version. Next, you'll select the tools this prompt should have access to.

Tools allow agents to perform actions. In this example, we enable three system tools:

• data_source_search

• transfer

• hangup

Inside the prompt text, clearly define the agent’s role as a support agent. Describe how it should answer questions, when it should look up information in the data source, and when it should transfer a call to a human. Explicitly referencing tools in the prompt helps ensure predictable behavior.

Click Save.

Step 2: Create the Greeting Message

The message defines the first thing a caller hears.

Create a new message and give it a consistent name so it is easy to find later. Keep the greeting short and friendly. For example:

You can also briefly explain what the agent can help with, such as policies, equipment, or general store information.

Click Save.

Step 3: Add a Data Source

Data sources store the information your agent uses to answer questions. This might include policies, products, contact information or hours of operation.

Create a new data source and name it consistently with your other resources. Add the information that you want the agent to reference. In this demo, the data is intentionally simple, but real-world agents often rely on much more detailed content.

Click Save.

Step 4: Choose a Voice and Language

Next, create a voice group. Voice groups define the language and voice your agent will use and can support multilingual experiences.  

Give the voice group a name, select a language, and choose a voice provider and then check out the voices. Each provider offers different voice styles. Use the preview feature to listen to samples and select a voice that fits your brand. Select the voice that's right for you.

Click Save.

Tip: You can select a new voice for each language.

Step 5: Assemble the Agent

With all resources in place, you can now create the agent.

Click on Agents, and then create a new agent and give it a name. Select the prompt, message and voice group you created earlier.  Under "Tools with configurations", verify that the correct data source is selected.

After saving, the agent opens in the preview window. Click Test and choose Start next to voice or chat. From here, you can immediately test the agent by speaking with it and observing how it responds, uses tools, and references data. Pay attention the the panel on the right for detailed information about tool calls.

Reviewing Conversations and Agent Behavior

All conversations with your agent are stored in the Sessions workspace. Each session includes:

• A full transcript

• An AI-generated summary

• An audio recording

• Tool usage and latency details

These records make it easy to understand how your agent behaves in real conversations and identify opportunities for improvement.

Connecting the Agent to a Phone Number

To make your agent available over the phone, you need to connect it to a voice channel.

Navigate to Channels and click the target tab. Create a new target. A target represents the phone number your agent will answer. You will also need to configure a fallback number that calls will transfer to if the agent cannot handle the request. You can use your personal phone number here.

Select your phone number and agent and then save the configuration. Your agent is now live. You can call the number immediately and interact with the agent just like a real customer. If requested, the agent can seamlessly transfer the call to a human.

Monitoring and Optimizing at Scale

Beyond individual calls, Syllable provides tools for long-term monitoring and optimization:

• Dashboards aggregate metrics across many conversations

• Insights automatically evaluate sessions and surface trends

• Campaigns enable outbound calling and proactive outreach

These features help teams move from experimentation to production while continuously improving agent performance.

Build AI Agents for Every Idea

This walkthrough shows how quickly you can build and deploy a fully functional AI voice agent using the Syllable Console. Whether you are creating a simple support agent or managing a fleet of production agents, the same building blocks and workflows apply.

For more advanced tutorials, SDK examples, and detailed documentation, visit the Syllable Docs or interact with our community on Reddit.

Want to follow along? Get a free account at Syllable.ai