A question is something you configure a voice agent to ask while it’s already on a call, on top of the open conversation it’s having. The agent asks it out loud, listens for the answer, and can write that answer onto a field in your workspace so it shows up next to the call in your library instead of staying buried in the transcript.
Why add questions to an agent
An agent already understands and responds to whatever a caller says, but a free-flowing conversation doesn’t reliably leave you with clean, structured data. Add a question when you need a specific piece of information out of every call, such as a membership number, an email address, or a yes/no confirmation, and you want it captured the same way every time instead of hoping the caller mentions it and the agent catches it.
Where to find them
Open the agent and go to its Details tab. The Questions section sits below the rest of the agent’s settings and lists every question currently configured on it.
Add a question
- Select Add question.
- Pick a question from the list, or select Create a custom question to define your own.
- Set whether it’s required and how the agent should handle a caller who doesn’t answer.
- Optionally map the answer to a field.
- Select Save question.
Picking an existing question attaches a ready-made one, built from a shared library your whole company draws from: it already has a display label, an answer type, and the exact question text the agent will say, all of which you can still override for this agent.
Creating a custom question instead asks you for a Name, a Description, and an Answer type, then saves it to your company’s own question library so you can reuse it on other agents without rebuilding it. Give it a display label and the question text before you save.
Set the answer type
Every question collects one specific kind of value. The answer type controls what the agent expects back, and includes Text, Email address, Number, Date, Time, Date and time, Yes or no, one of several options, and a web address.
Decide what happens if the caller doesn’t answer
Two settings on each question control this:
- Required turns on a Required label next to the question in the list, and means the agent must collect an answer before it can end the conversation.
- Max ask attempts limits how many times the agent tries. The first attempt counts as attempt 1, and the hard maximum is 3.
Once the agent has used up its attempts without an answer, If the user does not answer chooses what happens next:
- Move to next question continues the conversation and stops retrying this one.
- End conversation has the agent politely end the call instead.
The question list shows both settings together under each question, for example “2 attempts • Moves to next question on no response.”
Map an answer to a field
Each question can optionally point at a field in your workspace under Map to field. When a caller answers a mapped question, Speak writes that answer onto the field after the call ends, so it appears as structured data next to the recording instead of only inside the transcript. Leave it set to No field mapping if you just want the agent to ask the question without saving the answer anywhere structured. If the field you want doesn’t exist yet, you can create one from the same dropdown without leaving the panel.
For what a field can do once your questions are writing to it, see AI fields.
Edit or remove a question
Select the pencil icon next to a question to reopen it with its current settings, or the trash icon to remove it from the agent. Removing a question only detaches it from this agent; the underlying question stays in your company’s library so you can attach it to another agent later.
When you edit an existing question, the underlying question it’s based on is fixed. You can still change its required flag, attempts, no-response behavior, field mapping, and override its display label or question text for this agent.
What order questions run in
The agent asks questions in the order they appear in the list.
Related: Voice agent questions API · AI fields