Set up an inbound Voice AI agent

Set up an inbound Voice AI agent that handles incoming calls using natural speech. Instead of navigating traditional IVR menus, callers speak freely and the agent understands their intent, reasons about the appropriate response, and replies with natural-sounding voice.

Prerequisites [#prerequisites]

• Infobip account. If you do not have an account, sign up for an account.
• AgentOS enabled. Contact your dedicated account manager or Support if AgentOS is not available on your account.
• AI Agents Manager role assigned to your user.
• Voice channel enabled. Contact your Infobip account manager to enable it.
• A phone number registered with Infobip. Go to the Infobip web interface > Numbers section to get your voice number.

Products and channels [#products-channels]

• AI Agents
• Journeys
• Voice

Implementation [#implementation]

Step 1: Create and configure the AI agent

1. Log in to the Infobip web interface. Go to AI Agents, then select Agents.
2. Select Create agent. Choose From scratch or select a template. Templates such as Appointment booking agent pre-populate the instructions.
3. Give the agent a clear, descriptive name that reflects its purpose.
4. Write the agent instructions (system prompt). For voice-specific prompt guidance, see Prepare Voice AI agent for launch.
5. Add any tools the agent needs, such as Journeys components for backend integrations, MCP servers, or Exchange Marketplace.
6. (Optional) Attach a Knowledge Agent or knowledge sources for FAQ-style retrieval.
7. Test the agent in the Preview panel using the Chat and Traces tab.
8. Publish the agent. Only published agents are available in the IVR flow element.

Step 2: Set up the IVR flow

1. In the Infobip web interface, go to Automation Studio.
2. Create a new journey or open an existing one.
3. Add an IVR session to the flow.
4. Configure any initial routing steps. For example, add a greeting message or a DTMF routing step to direct callers to the right queue.

Step 3: Add the Transfer call to AI agent element

1. Add the Transfer call to AI agent element to the IVR flow where you want the AI agent to take over the call.
2. In the element side panel, select your published agent from the Choose AI agent list.
3. Select the Agent type:
   • Speech to speech: A single AI model listens and responds in voice directly, without converting to text. This enables faster responses and a more natural tone.
   • Speech to text to speech: Voice is first transcribed into text, processed by a language model, and then converted back into speech. This approach is more cost-effective and offers greater control.
4. Configure the remaining settings based on the selected agent type:
   • Speech to speech: Configure the voice, noise reduction, and VAD settings.
   • Speech to text to speech: Configure the greeting, language, voice, and transcription settings.

Step 4: Configure call session outcomes

Connect the exit branches of the Transfer call to AI agent element:

• Transfer to human agent requested: Connect to an Inbox queue or external contact center transfer.
• AI agent session ended: Add a closing message and a Hang up element.
• Call not transferred: Add fallback IVR handling or an error message.
• Call ended: Log the event. No further action is required.

Step 5: Test the call

1. Validate and launch the flow.
2. Call your configured phone number from a test device.
3. Verify the full call flow: greeting plays, agent responds naturally, tools execute correctly, session ends correctly.
4. Review the Chat and Traces tab in the AI Agents preview for debugging.

Next steps [#next-steps]

• Set up an outbound Voice AI agent call. See Set up an outbound Voice AI agent.
• Explore the AI Agents documentation.