Who it's for
Design partners, not customers, for now
Hospitals and health systems
Clinicians are already putting clinical questions to AI tools. Your safety and governance teams need a way to judge those answers on what they leave out, not only on whether their citations check out.
What a pilot could involve
- Bring the clinical questions your teams actually ask, and the omissions that would worry you most.
- Help us set the severity of missing points for your specialties.
- Review what the verifier flags, and what it misses, before anything goes near a ward.
Medical centers
Academic and community centers teach, audit and standardize how evidence reaches patients. An AI answer that is fluent but incomplete is a teaching problem and a quality problem at the same time.
What a pilot could involve
- Contribute cases from your specialties to the evaluation set, reviewed by your clinicians.
- Use flagged omissions in teaching and case review.
- Shape how completeness is reported to clinicians and quality teams.
Doctors and clinicians
You already check AI answers against what you know. The hard part is noticing what isn't there, especially outside your own specialty or at the end of a long shift.
What a pilot could involve
- Review answers and tell us which missing points matter clinically and which are noise.
- Test the interface: is a missing point clear, sourced and quick to act on?
- Help write the evaluation cases the verifier is held to.
Researchers
Public benchmarks score AI answers. Very little work scores a verifier: how often it catches a severe omission, and how often it cries wolf.
What a pilot could involve
- Work with us on the evaluation set and the metrics: severe-omission recall, precision and false-flag rate.
- Get access to the locked evaluation and its results as they are produced.
- Co-author or independently replicate the evaluation we publish.
Ground rules
What every pilot starts from
No patient records to start
Pilots begin with published cases and synthetic examples. Patient data comes in only once the contracts and technical safeguards it needs are in place with you.
Clinicians make the decisions
The verifier shows what an answer did not state. It does not tell anyone what to do, and nothing goes near a ward until your clinicians have reviewed what it flags.
Design-partner pilots
Tell us where omissions would hurt most
We are looking for a small number of hospitals, medical centers, clinicians and research groups to shape the verifier and its evaluation with us.
- ≥85%Target
of severe omissions caught
- ≤10%Target
of complete answers wrongly flagged