🔐 Who can use this feature?
Survey types: Attributed (Snapshot), Attributed (Continuous), Unattributed
Subscriptions: Any package with Engagement
Dynamic follow-up questions grounded in People Science
AI follow-ups adds a dynamic, AI-generated follow-up question to a rating or free-text survey question. When a participant answers, AI follow-ups can ask a personalized, in-the-moment question that probes the "why" behind what they selected or wrote.
The follow-up questions are grounded in People Science. They're designed using Socratic questioning (asking for clarification, then for reasoning or evidence) and appreciative inquiry (inviting what would improve a situation), so they stay neutral and don't lead the participant toward a particular answer.
Before AI follow-ups, the only way to ask a follow-up question based on how someone answered was Rating branching, a static, admin-written question attached to a specific answer. AI follow-ups fills the gap when you want to understand the "why" behind a response but can't predict in advance what that "why" will be, because it varies from person to person.
Check whether your account has AI follow-ups
AI follow-ups is switched on for each account by Culture Amp. It isn't something you can turn on yourself, and there's no sign-up form to fill in. You'll also need Engagement surveys on your account, because AI follow-ups only works on Engagement and Lifecycle surveys.
ℹ️ Note: AI follow-ups is rolling out in stages, so not every account with Engagement has it yet.
Today, AI follow-ups is available for English-language Engagement and Lifecycle surveys on a limited group of accounts. To find out whether yours is one of them, speak to your Culture Amp Customer Success Manager, or open a Support Conversation. If your account has it and a colleague can't see it in the survey designer, check that they have access to edit questions on that survey.
AI follow-ups vs. rating branching: which one should I use?
AI follow-ups and Rating branching solve related but different problems, and you can use both in the same survey:
Use Rating branching when you already know what follow-up question you want to ask, or you need structured, comparable data (for example, a rating or select-question follow-up you can compare across groups or over time).
Use AI follow-ups when you want to understand the "why" behind a response, but that "why" is likely to vary from person to person and you can't write a single follow-up question that covers it.
For guidance on designing static follow-up questions, see People science guidance for using rating branching and Best practices for using branching logic in surveys.
Which question types support AI follow-ups?
AI follow-ups can be enabled on:
Rating questions
Free-text questions
AI follow-ups can't be enabled on:
Select questions (single- and multiple-choice): AI follow-ups isn't available on these question types.
Demographic questions: These ask for facts about a person rather than an opinion, so there's nothing for a follow-up to explore. Following up here could also make individual participants easier to identify in your results.
eNPS questions (the 0 to 10 scale): AI follow-ups isn't available on eNPS questions.
See Question types used in surveys for more on how each question type works and its impact on reporting.
Enable AI follow-ups on a question
Navigate to Survey configuration > Questions
Locate the rating or free-text question you want to edit
Select the ellipsis to reveal the menu and choose Edit
In the question's settings, turn on AI follow-ups. AI follow-ups works best in English at the moment, and you'll see a note about this on screen.
If it's a rating question, use the Response field to choose which answers you want a follow-up question on (see below).
Set the maximum number of follow-up rounds you want to allow, in Maximum follow-ups (see below).
Save your changes.
Because AI follow-ups is enabled per question, you can turn it on for some questions in a survey and leave it off for others.
Turning the AI follow-ups toggle back off hides Maximum follow-ups and Response but keeps the values you chose, so turning the toggle on again restores your settings rather than resetting them.
Choose how many follow-up rounds to allow
You can set the maximum number of follow-up rounds a participant can go through for a question, to either 1 or 2. The Maximum follow-ups field defaults to 1.
Culture Amp's People Science team recommends setting this to 2, which is the highest value available. Their guidance is that the value of the answers you collect rises sharply from one round to two, and that going beyond two rounds would risk participant fatigue, which is why 2 is the highest setting.
As a general guideline, enable AI follow-ups on no more than about 3 questions per survey (roughly 5% of your total questions). Treat this as a ceiling, not a target, and count it together with any questions that already use Rating branching. Every additional question uses up a participant's limited attention, so it's worth reserving AI follow-ups for the questions where the "why" matters most.
Choose which answers get a follow-up question
For rating questions, use the Response field to choose which answers you want a follow-up question on. You can choose more than one answer here. Tick each answer on that question's rating scale, such as Strongly disagree or Disagree, that you want a follow-up question on.
You'll need to choose at least one answer before you can save. There's no single option that covers the whole scale, so if you want a follow-up on every answer, tick each one.
What you choose depends on what you want to learn:
Select just your lowest scale values if you want to focus on understanding unfavorable responses, useful when you want to understand what's driving a low score.
Select just your highest scale values if you want to focus on favorable responses, useful when you want to capture what "good" looks like, so you can reinforce or repeat it.
Select all the values for the broadest coverage, useful after a deliberate change, for topics where sentiment is likely to be mixed, for wellbeing questions, or the first time you run a new question and don't yet know what to expect.
Free-text questions have no Response field, because there are no set answers to choose between. Any answer a participant writes is eligible for a follow-up.
Key AI follow-up behaviors to plan for
AI follow-up responses are always optional
Participants are never required to answer an AI follow-up question, even if the question it came from is required. Participants can move on without answering the follow-up.
AI follow-ups and question-level comments
When AI follow-ups is turned on for a rating question, and the survey has question-level comments enabled, participants can also leave a regular comment on that question. Both are available at the same time: the comment box remains on, and the AI follow-up conversation runs separately.
Free-text questions keep their normal text response either way, since their answer is what the follow-up responds to.
What happens when an answer changes
If a participant changes their answer after an AI follow-ups conversation has started:
Rating questions: The follow-up conversation they'd had so far is removed. If their new answer is one you chose to follow up on, they'll get a new follow-up question. If it isn't, they won't see one.
Free-text questions: Editing their answer starts the conversation again, with a new follow-up question based on what they now wrote. Participants see a warning first, so they know that editing will restart the conversation.
If a participant edits an answer partway through: Everything after the answer they changed is removed. Changing their original answer to the question starts the whole conversation again, while changing their answer to the first follow-up keeps that first question and answer, and clears only what came after it.
What happens if a response is reset or a follow up doesn't generate?
If a participant resets their entire survey response, any AI follow-ups conversations they had are cleared along with the rest of their answers.
Occasionally a follow-up question can't be generated. The participant sees a short message in its place telling them to carry on with the survey, and they can still finish and submit as normal.
A participant can submit while the AI is still writing a follow-up question. They're never held up waiting for one to arrive.
Why a conversation can end early
A conversation can finish before the participant reaches the Maximum follow-ups you set, and often will. That happens when:
The participant dismisses the follow-up, or leaves it unanswered.
There's nothing more to explore in what they wrote. A brief reply along the lines of "all good thanks" may draw no follow-up at all.
If an answer touches on a sensitive topic, the AI stops rather than asking more about it. You'll see the participant's original answer, but no further questions.
An early finish isn't an error, and it never stops anyone completing the survey. Expect a mix in your results: some participants will have a full conversation, some will answer one question and stop, and some will get no follow-up at all.
If you turn AI follow-ups off after a survey has started
You can turn AI follow-ups off on a question part way through a live survey, and Culture Amp can turn it off for your whole account if you ask. Either way, any follow-up conversation a participant had started but not yet submitted won't reach your results. Conversations in responses already submitted are kept
Reading and interpreting AI follow-ups results
Read each conversation top to bottom, as one exchange, rather than pulling out single lines. A few things to keep in mind:
They're qualitative and directional, not representative. You'll only see answers from the people who got a follow-up and chose to reply. A theme that comes up repeatedly is worth investigating further, not proof of how common it is across everyone who took the survey.
Unanswered follow-ups don't reach your results. If a participant was asked a follow-up and never answered it, you won't see that unanswered question. You'll see the conversations people chose to have, not every question the AI offered.
They're not included in AI comment summaries. AI follow-ups conversations aren't included in AI comment summaries, which continue to summarize standard comments only.
They do appear in the comments report and comment export. You can review the full text of AI follow-ups conversations there, alongside standard comments.
A good starting point is to review your scores first, then read the conversations for the questions that matter most. Apply your usual filters as you go, and keep your minimum group size in mind so you don't read results for groups too small to report on. Use what you find to focus where you investigate further, rather than as proof on its own.
Tell the AI's questions from the participant's answers
In the comments report, a follow-up conversation appears below the comment it belongs to, separated by a line, reading top to bottom in the order it happened. Nothing labels which lines came from the AI, so it's the look of the text that tells you which is which:
The AI's questions are in smaller gray text.
The participant's answers are in the normal comment text style, the same as any other comment.
A purple Follow-up response tag sits at the bottom of any comment that contains a conversation, so you can spot them as you scroll.
So when you're reading a conversation, take the smaller gray lines as the AI's questions and the normal-sized lines as the participant's own words.
Filter your results to comments with a follow-up
The comments report has an Any follow-up responses filter with three options:
Show all: every comment, with and without a follow-up conversation.
With follow-up responses: only comments that have a follow-up conversation.
Without follow-up responses: only comments that don't.
The filter only appears when at least one question in the report has follow-ups enabled. For more on the other filters available, see Set Up a Filter for the Comments Report.
Export follow-up conversations
Follow-up conversations are included in two exports:
Your comments report export now includes an AI Follow Ups column containing the conversation. If you're exporting comment translations too, you'll also get AI Follow Ups English and AI Follow Ups Detected Language columns.
Your Raw Data Extract now includes one extra column for each question that has follow-ups turned on. Each one is named after the question, ending in .ai-followups, and you'll find it grouped with that question's other columns, next to its comment and tag columns. It appears only when you run the extract with Include questions comments ticked.
💡 Limitations
English works best: AI follow-ups is designed for English. If you've translated your survey into other languages, participants may still get follow-up questions, written in the language they answered in, but we haven't tested those languages yet so the questions may not be as good. In the comments report, you can only translate a follow-up conversation into English.
Rating and free-text questions only: Select questions, demographic self-select questions, and the eNPS 11-point scale question aren't supported.
Not available in AI comment summaries: AI follow-ups conversations aren't included in AI comment summaries but do appear in the comments report and comment export.
Participants can always finish the survey: Even if a follow-up question doesn't appear, participants can complete and submit their response as normal.
FAQ
The AI asked about something I didn't expect. Is that a problem?
The AI asked about something I didn't expect. Is that a problem?
Follow-up questions respond to what the participant actually wrote or selected, so the exact wording varies. The AI is designed to avoid asking about sensitive topics, but that doesn't replace your own checks. You'll still want to review and redact comments the way you normally do.
Why do some comments show AI-generated questions alongside the participant's answer?
Why do some comments show AI-generated questions alongside the participant's answer?
AI follow-ups conversations display the AI's questions next to the participant's answers so you have context for what they were responding to. Sentiment and topics are based only on what the participant wrote, not on the AI's questions.
Do AI follow-ups support the eNPS 11-point scale or select questions?
Do AI follow-ups support the eNPS 11-point scale or select questions?
Neither is currently supported. AI follow-ups can only be enabled on rating and free-text questions.
➡️ Next steps
People Science Guidance for Using AI Follow-Ups: choose which questions to enable and how to read what comes back.
Design and configure survey questions: set up the rating and free-text questions you'll enable AI follow-ups on.
Question-level comments in surveys: benefits and drawbacks: decide whether to keep the comment box on alongside a follow-up conversation.
Understanding Raw Data Extracts: work with the new follow-up column in your extract.
💬 Need help? Open a Support conversation to speak with our Product Support team.