Built on the OpenExO model
OpenExO is an operating model for exponential organizations: a Massive Transformative Purpose (MTP) plus an Intelligence Stack that senses, interprets, decides, acts and learns—wrapped in a permanent governance layer.
We adopt it as our operating system, then add the clinical assurance layer OpenExO does not provide. OpenExO supplies the organizational skeleton; ProvidEHR supplies the clinical control plane; InVivo supplies the personal edge; the outcome ledger supplies the learning engine. Critically, we compile the MTP into a machine-readable constitution every agent and workflow must obey.
“Make early prevention continuous and available to everyone.”
OpenExO's central bet is that most workflows will become agentic. The responsible move is to build the architecture and governance for that near future now—so autonomy can arrive under control, not as an afterthought.
ProvidEHR is built for that world from day one: agents do the coordination, every consequential action passes a machine-readable constitution and an autonomy gate, humans move onto the exception and accountability paths, and authority only widens as measured outcomes earn it.
The Intelligence Stack
Six operating layers, each wired to a component we ship.
A machine-readable Prevention Constitution: hard constraints + weighted priorities, enforced before every consequential action.
InVivo, wearables, labs, EHR data, questionnaires and population registries.
Longitudinal state, risk and control states, uncertainty, missingness and applicability.
Eligibility, evidence-based protocols, benefit–harm estimation and autonomy tiers.
CarePlans, outreach, scheduling and clinical queues—staged for review, never an autonomous clinical write.
Prediction → intervention → adherence → outcome, joined into calibration, equity and drift.
An always-on control plane: computable consent and purpose limitation, autonomy tiers A0–A3, clinician attestation, audit and provenance, drift and equity monitoring, break-glass and kill-switches.
Purpose, made executable
The MTP isn't a slogan—it's a constitution enforced before every consequential write: seven absolute 'never' rules plus weighted priorities (safety > speed, equity > ease, reversibility > autonomy).
Measured, not assumed
Every prediction is joined to what actually happened—with censoring—into calibration, per-subgroup equity and drift, closing the compounding loop.
Autonomy is earned
A model's clearance comes from its measured performance and fails closed to draft-only until it is calibrated, stable and equitable. Order and prescribe autonomy stays human.
Prove beside the incumbent
One complete governed pathway—home blood-pressure detection and follow-up—runs next to the existing EHR and migrates once it demonstrably helps.
Where we don't follow OpenExO literally
- Clinicians stay on the clinical-accountability path—agents coordinate, they never diagnose or prescribe.
- Escalation is never suppressed for cost or capacity.
- Autonomy is earned through outcome and safety evidence, not technical capability alone.
- We optimise for healthy-life improvement per unit of patient burden and clinical capacity—not engagement or growth.