Evidence before claims
Journai Health is committed to evaluating product performance, usability, and safety before making claims about clinical or operational impact.
Last reviewed: 1 August 2026
- We do not publish clinical impact without validated data.
- Known gaps and uncertainties are documented openly.
- Evaluation protocols are developed with clinical input.
- Success measures are agreed with each pilot organisation.
- Clinical value must be demonstrated, not assumed.
Evaluation status
Current validation activity across testing, clinical review, and summary-quality assessment.
| Evaluation area | Method | Progress |
|---|---|---|
| In progress | ||
| Synthetic-data testing | Controlled testing with synthetic FHIR datasets | Internal test results, not yet published Next step: Document methodology and publish summary on this page Updated 2026-08-01 |
| Summary-quality evaluation | Structured assessment of summary accuracy and completeness | Internal evaluation framework, not yet published Next step: Complete initial evaluation cycle and publish methodology Updated 2026-08-01 |
| Hallucination testing | Systematic testing for fabricated or unsupported clinical content | Initial test approach documented internally Next step: Publish test methodology and aggregate results Updated 2026-08-01 |
| Source-attribution testing | Verification that summary statements link to correct source records | Internal testing underway Next step: Publish attribution accuracy methodology Updated 2026-08-01 |
| Evaluation planned | ||
| Usability testing | Structured sessions with clinical and operational stakeholders | Evidence not yet published Next step: Conduct initial usability sessions with pilot partners Updated 2026-08-01 |
| Clinical review | Clinician review of AI-assisted summaries against source records | Evidence not yet published Next step: Define review protocol with clinical advisors Updated 2026-08-01 |
| Omission testing | Assessment of whether key clinical information is missed | Evidence not yet published Next step: Define omission test cases with clinical input Updated 2026-08-01 |
In progress
- Synthetic-data testing
- Controlled testing with synthetic FHIR datasets
- Internal test results, not yet published Next step: Document methodology and publish summary on this page
- Updated 2026-08-01
- Summary-quality evaluation
- Structured assessment of summary accuracy and completeness
- Internal evaluation framework, not yet published Next step: Complete initial evaluation cycle and publish methodology
- Updated 2026-08-01
- Hallucination testing
- Systematic testing for fabricated or unsupported clinical content
- Initial test approach documented internally Next step: Publish test methodology and aggregate results
- Updated 2026-08-01
- Source-attribution testing
- Verification that summary statements link to correct source records
- Internal testing underway Next step: Publish attribution accuracy methodology
- Updated 2026-08-01
Evaluation planned
- Usability testing
- Structured sessions with clinical and operational stakeholders
- Evidence not yet published Next step: Conduct initial usability sessions with pilot partners
- Updated 2026-08-01
- Clinical review
- Clinician review of AI-assisted summaries against source records
- Evidence not yet published Next step: Define review protocol with clinical advisors
- Updated 2026-08-01
- Omission testing
- Assessment of whether key clinical information is missed
- Evidence not yet published Next step: Define omission test cases with clinical input
- Updated 2026-08-01
Research and publications: Evidence not yet published.
Known limitations
Constraints and uncertainties that should be understood when reviewing the platform and its outputs.
- AI-assisted outputs may contain errors or omissions and require clinician review.
- Summary quality depends on completeness and quality of source data.
- Not all clinical workflows or systems are supported in the current MVP.
- Regulatory classification and compliance status may change as capabilities evolve.
- No clinical benefit should be assumed until evaluated with agreed metrics.
Discuss a clinical information workflow.
Speak with Journai Health about evaluating the platform within a defined workflow in your organisation.