The VoicePulse field guide
AI voice surveys: a practical guide to better feedback.
A useful conversation starts before you press record. Here’s how to choose your goal, write the questions, and turn spoken feedback into something you can work with.
By the VoicePulse team · Updated
What is an AI voice survey?
An AI voice survey collects spoken feedback through a conversation with an AI interviewer. Unlike a fixed form, it can ask follow-up questions to clarify an answer. Teams use the resulting transcripts to understand experiences, needs, and decisions.
The interviewer follows instructions you provide. A participant speaks into their microphone, and the conversation can explore an answer in more detail. The result is a transcript you can review alongside your research notes.
VoicePulse applies this approach to customer feedback, user research, and candidate screening. A reusable interviewer is called a Voice Pulse; each participant’s session is a conversation.
When should you use voice instead of a form?
Use a voice survey when the explanation matters: onboarding friction, a product decision, or a recent experience. Use a form when you need short standardized answers, precise numeric inputs, or a quiet and accessible text-based option.
Good reasons to open a conversation
- You know a customer struggled, but you do not know which part of the experience caused it.
- You want someone to describe a recent workflow in their own words.
- You need examples and context that a multiple-choice answer would leave out.
When another format is a better fit
A form is often simpler for ratings, dates, and standardized inputs. A human interview is valuable when the topic is sensitive or requires a researcher’s judgment. Do not assume everyone wants to speak: provide an alternative for participants who cannot use a microphone or prefer to type.
What makes a good voice survey question?
Ask about one recent, concrete experience in neutral language. For example: ‘Tell me about the last time you tried to set up your account.’ Follow with a request for context rather than suggesting the answer you hope to hear.
- Start with a specific experience. “Tell me about the last time you used this feature.”
- Ask what happened. “What did you do after you reached that step?”
- Explore the difficulty without suggesting one. “What, if anything, was unclear?”
- Follow the participant’s words. “You mentioned a workaround. Can you walk me through it?”
- Leave room for what you missed. “Is there anything else about the experience you’d like us to know?”
Avoid packing several questions into one sentence or telling people what a “good” answer looks like. For more starting points, explore the conversation template library.
How do you plan your first AI voice survey?
- Choose one decision. Write down what you need to understand and how the answers could change your next step.
- Choose the right participants. Invite people with relevant, recent experience. A convenient group is not necessarily representative of everyone you serve.
- Draft a short conversation. Start with a small set of open questions. Include instructions about what the interviewer should explore and what it should avoid.
- Test the complete experience. Open the link, check microphone permissions, answer in different ways, and review the transcript. Adjust confusing wording before inviting participants.
- Explain what to expect. Tell participants they will speak with an AI interviewer, what the feedback is for, and how their responses will be handled.
- Review and iterate. Check the first responses before expanding the invitation. Revise questions that produce vague or off-topic answers.
In VoicePulse, create a Voice Pulse in the dashboard, customize its instructions, and share its conversation link. Start with the free plan to test the workflow before choosing more capacity.
How should you review the responses?
Read the transcript before jumping to a conclusion. Keep a participant’s quote alongside your interpretation so another person can understand the context. Group recurring observations, then look for answers that disagree with the pattern.
Spoken feedback explains experiences; it does not automatically tell you how common an experience is. A small, self-selected group can reveal useful questions without representing your entire customer base. AI transcripts and summaries can also contain errors, so verify important details before making a decision.
Finish with a specific next step: clarify onboarding instructions, investigate a recurring workaround, or schedule a deeper human interview. The point is to make your next question or decision better informed.
Further reading
The UK Government Service Manual’s guide to in-depth interviews covers planning interviews and asking questions that help people describe their own experiences.