The future of survey programming
is already shipping.
Insights runs on surveys. Surveys run on code — until now, hand-authored by specialists, measured in days. Fieldbase SP writes that code the engine was built to run, inside the editor the team already uses, with every change in a thread the operation can audit. A 60-question tracker, drafted in minutes.
Drop the brief.
Get the survey.
A client doc arrives in the morning. By lunch, a working draft is on a reviewer's screen — not next week.
- Word, PDF, or transcript — in. Fieldbase SP parses the structure, keeps the sections, keeps the ordering.
- Full survey — out. Screeners, routing, quota wiring, options, scripts. All linted before you see it.
- Level 3 complexity · 60 q · ~6 min. A survey a specialist would quote for 24 hours, drafted while you refill a coffee.
if S2.value >= 65: if S2.value < 18 or S2.value >= 65:Describe the change.
The AI writes it.
You accept, refine, or skip.
Every other AI you've used lives in another tab. You copy, paste, hope. Fieldbase SP edits the project in place — and nothing lands without a human click.
- Ask inside the editor. Scope with
@,qid, or a line number. No copy-paste tab-dance. - Review the diff. Lint-checked before you see it. Accept, Refine, or Reject with one click.
- Lands in place. No stitching, no "did it get the quota cell too?" — the editor is where change happens.
Every change,
in one audit thread.
The team sees everything. Scoped to one project. Scroll back weeks. Ask the AI itself what changed last Thursday — it answers from the log.
- Persistent, per-project. Not a personal chat window. Each project has its own thread, bound to its own file.
- Shared with the whole team. Every ask, every proposal, every Accept, Refine, Reject, revert — logged where everyone can see it.
- Ask the thread itself. "Summary of changes this week?" The AI answers from its own log — the audit trail nobody has to maintain.
- ✓ S1b Gender added (Priya, Tue)
- ✗ QMaxDiff proposed, reverted (Sarah, Tue)
- ✓ S1 → COUNTRY renamed, 13 refs updated (Dev, Thu)
Still a human decision
The AI never writes directly to the survey. Every proposal is a reviewable diff with Accept, Refine, or Reject. Metering is opt-in; a team can leave AI assist switched off and nothing else in the editor changes. What ships is the research team's call — the AI just takes the slow part out of the middle.