Simulation
How it works
Real interviews in.
A persona you can question out.
One participant becomes a whole population you can interview, any time — grounded only in evidence a researcher has reviewed and approved.
For researchers
Intake
Collect real evidence. Upload interview transcripts (PDF, DOCX, text) or invite participants to an AI interview — by voice or text. Everything is parsed into verbatim evidence chunks, the raw material the rest is grounded in.
Source & approve
The system proposes public sources that could enrich the persona. Nothing is ingested automatically — you approve, reject, or defer each one. That gate is the trust boundary.
Synthesize
Generate a persona from the approved evidence: a coherent character, a contradiction map where the evidence disagrees with itself, and full provenance back to every source.
Talk
Interview the persona by text or voice — responses stay tethered to the evidence, and every session is saved and summarizable.
Panels
Run a sampled synthetic audience against a concept, message, or price — reactions mapped to rating distributions you can calibrate against real survey data.
Focus groups
Several evidence-grounded agents discuss your questions over rounds, reacting to each other, then a cited readout of themes and tensions.
Knowledge graph & briefing
An entity/relationship graph over the evidence powers sharper retrieval, surfaced tensions, and a multi-perspective, self-graded research briefing.
For participants
If you've been invited to a study, your job is simple: have a conversation. Here's what to expect.
Get invited
You receive an invite (by email or a link) to a specific research project.
Sign in
Use the same email you were invited with — a passwordless link works, no password to remember.
See only your projects
Your portal shows just the studies you've been invited to, nothing else.
Have the conversation
Answer in whichever way suits you — speak out loud or type. Plan for ~15–20 minutes. Mute or end whenever you like.