It is a workspace element, like the chat agent, but it talks instead of writing. It knows what you write in its instructions, the prompt. You set it up in four steps and test it with a Web Call from your browser, even before you have a number.
#What it can do
- Answer or call. An “Inbound” agent answers people who call its number. An “Outbound” agent calls your contacts, from a campaign or an automation. You choose this in the Agent function field.
- Greet callers your way, with a message that can change with the time of day or with the customer.
- Follow your instructions: answer questions, collect and confirm name, phone and email.
- Hand the call to a teammate and book appointments on Cal.com, Calendly or Google Calendar.
- Work after the call: it sends you an email with the outcome, a summary and the extracted data, passes them to your business software and can start an automation.
#How it connects to the phone
The agent answers on the numbers you assign to it: how to get them is explained in Connecting your phone: the options. If you need both directions, create two agents: they can use the same number, as explained in One number, several agents.
#Where to find it in the console
Voice agents are in the workspace Overview, in the “Voice Agents” panel, each with its own card. When you open an agent, the sidebar shows the “Voice agent” group.
| Menu item | What you find |
|---|---|
| Dashboard | The getting-started guide and the call statistics. See Dashboard. |
| Calls | The list of calls, with recording, transcript and summary. See Calls: list and details. |
| Agent profile | All the agent's settings, divided into four steps. |
#The top bar
On every agent page, in the top right, you find three controls.
- The number. You see the agent's first number, with “+N” if it has others; click it to manage them. If the agent has no numbers, you see Assign number. See Assigning and managing numbers.
- Test Call, with two tests: Phone Call and Web Call. We explain them in Testing the agent.
- The status, Active or Paused. Click it to change it.
#The Agent profile: summary and four steps
The Agent profile item opens the “Voice Agent Profile” page, a summary with one area for each of the four steps. In the top right you find the agent's direction, “Inbound” or “Outbound”, and the Edit Profile button. You only see the areas your permissions give you access to.
Each area has the Set up or Edit button and a status label. “Configured” means the area is complete. “Partial”, “To configure” and “Needs fixing” mean something is left to complete: open the area and look for the flagged field. “Not set” means optional features that are all off.
| Step | What you set | Where to read |
|---|---|---|
| 1. General | Name, language, voice, greeting, call timings, recording, advanced settings. | Identity, voice and greeting · Behavior and advanced · Recording and consent |
| 2. Prompt & Instructions | The text the agent follows on every call. | Prompt & Instructions |
| 3. In-call | Transferring to a teammate and booking appointments. | Transfer · Appointments |
| 4. Post-call | Data to extract, summary email, webhook. | Post-call |
#Saving your changes
Edit Profile, or an area's Set up, opens the “Configure Voice Agent” page, with the four steps at the top: you can open them in any order.
The save bar is always at the bottom. Save changes saves the changes in all steps together and Cancel discards them. Next step saves and moves to the next step; on the last step, Finish saves and returns to the summary. If a field is wrong, the page takes you to the field to fix, even if it is in another step.
Save before opening another page. While you switch steps, your changes stay pending; if you leave the page, they are lost.
What doesn't go through Save changes. Some things save on their own, or aren't saved with the bar:
- the guided generation fields: they are saved only with Generate Prompt, when the generation finishes, together with the new prompt. See Prompt & Instructions;
- the agent's numbers: adding, changing or unlinking them takes effect immediately, from their own windows. See Assigning and managing numbers;
- the webhook secret key: Regenerate the key changes it immediately, after you confirm;
- the status Active or Paused: you change it from the top bar and it takes effect immediately.
On the other hand, window buttons such as Save voice, Save Conditions, Add destination or Link event only confirm the window: to save, you still need Save changes.
#The recommended cycle
- Create the agent and write the prompt
- Check the voice and greeting
- Test with a Web Call
Correct the prompt until the agent answers the way you want.
- Assign a number and add the actions
Transfer, appointments, email and webhook. Test them with a real call.
- Listen and improve
In the first weeks, read the calls and update the prompt with the questions the agent couldn't handle.
#Active, paused, forced pause
You change the status from the label at the top, and it requires the “Create and delete agents” permission.
| Status | Inbound calls | Outbound calls |
|---|---|---|
| Active | The agent answers. | They go out. |
| Paused | They are rejected. They stay in Calls as “Unsuccessful”, with “Agent paused at call time”. | They don't go out. |
| Forced pause | Your plan no longer covers the agent: it is stopped, as if paused. See The forced pause. | |
While paused, you can still test the agent with a Web Call. If “Your plan does not allow activating more agents” appears when you reactivate it, you have reached your plan's limit. The other reasons why a call can be rejected are in If the call is rejected.