Governed AI · healthcare decisioning
PriorAuth Sentinel
The governed inverse of nH Predict. A prior-authorization engine where deterministic code decides and the model only locates and phrases. It cannot issue a denial without a specific, cited unmet criterion; when the chart is silent it abstains to a clinician instead of guessing; and every determination ships a replayable appeal packet. Pick a request, then flip the flagship case between a black box and the governed engine on the same chart.
⚠ synthetic data No PHI. Synthetic InterQual-style ruleset and seeded charts, this demonstrates a control architecture, not clinical correctness. Runs entirely in your browser.
Auth requests
·The structural guarantee. A criterion only becomes unmet when the chart contains an explicit contradicting fact. Absence of evidence is unknown, never unmet, so a denial can never be manufactured from silence.