DEMO. An unaffiliated design study built by Verzi for Posognos. Lab pages are concepts under exploration, not part of the proposed site. Every number below is cited at the foot of the page.
PsiBench edition 2026.1models 40evaluations 59,040last run 21 to 23 April 2026register 40 ratings · all current
PosognosMedication-safety evaluation for clinical AI
For health systems
6 hours.
Your hospital’s entire annual medication-safety evidence: one prescriber, two hard-timed windows, once a year, with a 120-day lockout before a retake.
The same drug knowledge base: as its vendor designed it, and as actually deployed across 358 hospitals. A third of the safety you bought is lost to local configuration, and it drifts every week.
of harm events among Medicare inpatients are medication-related. The single largest category of preventable harm in the building.
HHS Office of Inspector General, 2022 [3]
What continuous looks like
59,040
evaluations against 492 expert-authored scenarios, repeatable in days. No clinical time, no PHI, no production system. Run it before the scored attempt, and on every configuration change after.
PsiBench edition 2026.1, methods preprint [4]
The clock that started
1 April 2026
Leapfrog now requires every hospital to answer for the AI in its medication management. RUAIH, SAFER attestations and six state AI laws follow within a year. The evidence is your job now. Nobody handed you an instrument.
2026 Leapfrog Hospital Survey, Summary of Changes [5]
Continuous beats annual. Measured beats asserted.
The full argument, with every number sourced and the sample reports to file, is one click away.