Multi-stepClients
Turn individual answers into a shared understanding.
Collect structured feedback, explore ratings and themes, and ask your connected AI to publish a report from authorized responses.
A complete worked example
Turn consistent questions into an evidence-led report.
The product team wants to learn why new users hesitate during setup. Participants answer the same short questionnaire.
Fictional project and sample data. This preview does not submit, publish, or change access.
Product team · illustrative result
Where setup becomes unclear
A small qualitative sample gives the team questions to investigate, not a population-wide estimate.
- Sample
- 6 completed responses
- Strongest theme
- Unclear first action
| Theme | Mentions | Evidence |
|---|---|---|
| First action unclear | 4 of 6 | R01, R02, R04, R06 |
| Terminology unfamiliar | 2 of 6 | R03, R04 |
| More examples wanted | 1 of 6 | R05 |
Interpretation
Try a clearer first-action label and a short worked example. Keep R05's request visible as a useful suggestion without presenting it as the majority view.
What the recipient can access
An anonymized summary with references available to authorized researchers.
Kept out of this handoff
Participant contact details and unredacted answers.
How it runs
- Ask consistent questionsCombine ratings with open-ended questions so the team can compare responses without losing context.
- Explore the evidenceFilter responses by topic and chart ratings before asking your AI to summarize the findings.
- Publish a useful reportHave your connected AI query authorized records, cite response IDs, and publish a report with the same access restrictions.
- You getA consistent feedback form, a response table you can query, and a report in the same restricted project folder.
A prompt to start with
Copy this into your AI tool after connecting it to Fermium. Replace the bracketed details and let it ask for the context it needs.
Create a feedback form with participant reference, topic, rating, what worked, and what to improve. Keep responses in our restricted Research folder. Add a chart of ratings by topic. When I ask for a synthesis, use MCP to query completed responses, separate recurring themes from individual comments, cite record IDs, and publish an HTML report in the same folder. Do not include identifying information unless I explicitly request it.
What this workflow needs
AI synthesis runs through your connected AI tool when you request it. Access to the report should be reviewed before sharing, especially if it contains quotations or personal information.
Can the AI analyze only the new responses?
Yes. Save the database change cursor after processing, then request changes after that cursor on the next run. Fetch affected records under the same user's permissions.