← Back to V20
01 — Create

AI copilot builds your survey with you

Describe the outcome you need. Survona drafts the questionnaire, flags bias, translates into 40+ languages, and generates a live preview you can edit by talking to it.

Builder v2 — drag, drop, debug

A questionnaire canvas with full branching logic, validation, and a live debugger that replays respondent paths as you design.
  • 14 built-in app types (survey, CARM, pulse, baseline…)
  • Branching + skip logic with visual debugger
  • Translation to 40+ languages in one click
  • Templates library and version history

Authoring copilot

GPT-5.4 primary · Claude Sonnet 4.5 fallback. Streaming diffable patches so you always see what changed.
  • Bias + leading-question detection
  • Tone + reading-level control
  • Inline suggestions with citations
  • Audit trail of every AI edit

Agentery 8-stage pipeline

3-agent consensus (drafter · critic · judge) runs validation before any respondent sees the survey.
  • Stage 1–3 · plan, draft, translate
  • Stage 4–6 · critique, score, repair
  • Stage 7–8 · pretest with personas, publish
  • Full trace viewer per run

Live UI edits (design track)

Claude Opus 4.7 + Claude Design generate and patch UI screens in real time — scoped to design only, never business logic.
  • ScreenSpec JSON-Patch format
  • Design tokens locked to the Survona system
  • Human approval before merge
  • Launched 16–17 Apr 2026