Persona Simulations
Run a cohort of synthetic personas against one of a project’s prototypes, or against a question, and read back findings at population level. It is the fast way to stress-test an idea and generate hypotheses — hours instead of a fielded study.
Results are simulated. This is a way to form hypotheses quickly, not a substitute for evidence from real users. Treat findings as directional and confirm the ones you are going to act on. We would rather say that here than have you discover it in a board meeting.
Each persona reads a scenario or a question and reacts once. Good for opinion, intent and first impressions — the things a questionnaire would ask.
Each persona holds a multi-turn conversation with the target. Good for flows and assistants, where the interesting failure happens on turn four rather than turn one.
Filter by region, age, language, economic motivation, trust and tech savviness — and Preview cohort tells you how many personas match before you spend anything. Save it, and it is shared across every project in the account, so “our mid-trust European buyers” means the same population next month as it does today.
Add metrics that score each trial from 0 to 1, as pass, flag or fail. Use the built-in ones or write a custom rubric with your own thresholds. The run then reports each metric’s split rather than one blended number, because a blended number hides which part failed.
Open a run for a live view: status, a persona swarm that fills in as trials finish, each metric’s pass/flag/fail split, and a Findings section that ranks which traits — age, region, trust, income — genuinely shifted the result.
Statistically weak differences are flagged as weak. A simulation will always produce a number for every cut of the data; the honest part is saying which of those numbers you should not believe. Cancel a run at any point and keep whatever finished.
The setup screen shows an estimated cost before you commit, there is a per-simulation spend cap set by your platform administrator, and every run is metered to the token like every other AI surface in the Hub. Not sure how to set it up? Describe the question and Suggest setup fills the form in for you to correct.
Tell us what you are working on and we will open an account.
Request access