Survey Intelligence

1 Pick a survey

2 Filter respondents

Only submitted, non-excluded respondents are included. Optionally narrow by submit date.

From

include all dates

3 Choose questions

Generate questions the AI can confidently answer from existing responses, and/or write your own. Tick the ones to include. Click “Generate with AI” again to add more.

4 Preview — sample of 10

5 Generate for all respondents

Runs the included questions across the filtered cohort (capped for cost) and produces a raw-data Excel: every respondent with their real answers + the AI-generated answers.

1 Choose data source

The file should include a status column (rows with submitted) and an exclude column. Other columns become question labels.

2 Configure the run

Time scope multiple waves detected

From

include all dates

Select waves to compare (2 or more)

Tick 2 or more waves to compare.

Choose how to build the segments. Filters are applied automatically: status = submitted, exclude empty.

The AI picks the segmentation axes. Preview its plan and review it before running:

Additional details (optional)

The AI analyst writes & runs its own analysis — this takes a few minutes.

3 Results

⬇ Download PowerPoint Open report ↗

1 Pick a survey

Backtest the AI: feed it a subset of each respondent's real answers, let it predict the rest, then compare the predictions to what they actually answered.

2 Filter respondents

Only submitted, non-excluded respondents are included. Optionally narrow by submit date.

From

include all dates

3 Choose questions

A Questions fed to the AI (the persona)

Tick the questions whose real answers the AI is given to build the persona.

B Questions the AI should NOT answer

These are the remaining (held-out) questions — your seed questions are already excluded. Tick any you do not want the AI to answer. Whatever stays unticked is predicted and scored against the real answers.

4 Preview — sample of 10

5 Run for all respondents

Runs the backtest across the filtered cohort (capped for cost) and produces an Excel: every respondent's seed answers, real vs AI-predicted held-out answers, and a match column with per-respondent accuracy.