Multi-stepClients
Keep the findings close to the evidence.
Collect structured feedback and screenshots, then turn responses into a readable research report.
A complete worked example
Make detailed research readable for a busy project team.
A researcher has six interviews, a short feedback form, and annotated screenshots. Decision-makers need the themes, exceptions, and evidence in one place.
Fictional project and sample data. This preview does not submit, publish, or change access.
Restricted project readout · illustrative result
The first action needs a clearer label.
Four of six participants hesitated before starting. The pattern is consistent enough to test a revised label, not enough to predict conversion impact.
- Sample
- 6 interviews and their feedback records
- Recurring issue
- Unclear starting point
- Counterexample
- Two experienced participants completed setup easily
Evidence readers can follow
R01, R02, R04, and R06 describe uncertainty at the first action. Their supplied screenshots show where they paused. R03 and R05 already knew the workflow, which may explain their different experience.
Next study
Test a clearer action label with new participants. Use the same task and questions so the next readout can compare observations without inventing a causal result.
What the recipient can access
Anonymized findings and permitted evidence references.
Kept out of this handoff
Participant contact details, private transcripts, and unredacted screenshots.
How it runs
- Gather consistent responsesUse a form to capture observations and optional evidence.
- Interpret the materialAsk a connected agent to identify themes and cite the responses it used.
- Publish the readoutShare the report with the permitted audience and preserve source references.
- You getA sourced readout with themes, supporting examples, and open questions.
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 for [research question] with optional screenshots. After responses arrive, use the material you can access to identify themes and exceptions, cite the supporting records, and publish a readout for [audience]. Separate evidence from interpretation.
What this workflow needs
Interpretation needs an active agent with tools suited to the supplied material. Binary OCR and transcription are separate capabilities. Respect participant access and sharing requirements.
Are the themes calculated automatically when someone responds?
No. Run the analysis when requested, or configure an external runner for a recurring process.