AI follow-up questions
The form asks one smart clarifying question when an answer is too thin — you define what a complete answer contains in plain language.
This is the “forms that think” part. On any text field, tell the assistant what a complete answer contains — “should mention budget and timeline” — and the form probes for what’s missing when a respondent answers thinly.
How it behaves:
- One specific question at a time, generated from the actual answer, up to your
max_followups(1–3). - Attributed and optional — respondents see it’s AI and can ignore it.
- In multistep layouts, follow-ups render inline under the field or as their own “one more thing” step after Next (placement is your choice).
- Capped and degradable — token and volume ceilings per plan; when hit, the form silently becomes a static form. See limits.
The follow-up transcript is stored with each response, and the built-in A/B holdout tells you whether follow-ups actually improve your data.
Can respondents skip the follow-up?
Yes, always. Follow-ups are clearly attributed as AI-generated and optional — the submit path is never blocked.
What happens if the AI is unavailable or a cap is hit?
The form degrades to a static form. Respondents never wait on the AI layer; caps exist per response, per day and per month.
Which fields support follow-ups?
Text fields only (short_text and long_text). Enable them per field, up to 3 follow-ups each.