Trust
How Ozwell handles the record, and where a clinician stays in control
Ozwell drafts; a clinician decides. This page collects what we publish about how that boundary is enforced, how the AI is governed, and what we have had independently certified.
Controls
What is in place today
Each of these is described elsewhere on this site — in the product, the Help Center, or the certification disclosure. They are collected here so they can be read in one place.
Nothing files without approval
Ozwell drafts the note and queues orders, referrals, and follow-ups as reviewable actions. A clinician approves them before anything is written to the patient record, so a person is always the last step before the chart.
Designed to augment, not replace
The published intended decision-making role is to inform and augment clinical decision-making, not to replace clinical management. That is a disclosed design constraint, not a disclaimer added afterwards.
HIPAA compliant
Ozwell is built for use with protected health information in outpatient clinics, hospitals, and administrative offices.
Independently certified for AI risk
Drummond pDSI-Risk certification, under ASTP/ONC § 170.315(b)(11) — covering intervention risk management and source attribute disclosure. Ozwell was the first AI-powered health IT product to hold it.
Limits published, not buried
The settings and populations Ozwell should not be used in are published in full, including emergency and critical care and highly specialized care. So are the known limitations of its fairness and validity testing.
Risk management practices on the record
Our intervention risk management practices — how risk is identified, assessed, mitigated, and reviewed as the product changes — are published rather than described on request.
The approval gate
A clinician is always the last step before the chart
This is the single most important thing to understand about how Ozwell works, so it is worth stating on its own rather than as one bullet among six.
Ozwell listens to the encounter, drafts a structured note, and queues the orders, referrals, and follow-ups it identified. All of it arrives as a reviewable action. Nothing is written to the patient record until a clinician reads it and approves it.
That is a design constraint, not a setting. Our published intended decision-making role under ONC’s source attribute requirements is to inform and augment decision-making, not to replace clinical management — which means the boundary is part of what was independently reviewed during certification, not a preference we could quietly change.
Where Ozwell should not be used
The limits are published, not buried
A tool that never states its boundaries is asking you to discover them in production. Ours are part of the certified disclosure.
- Emergency or critical care settings where real-time clinical decision-making is required.
- As a substitute for professional medical judgment or expertise.
- Highly specialized or nuanced care, such as advanced pain management or rare disease treatment.
- Self-treating complex medical conditions without professional oversight.
The known risks and limitations behind these, including the limits of our fairness and validity testing, are set out in the source attribute disclosure and our intervention risk management practices.
Questions we have not answered here
Ask us directly
For security reviews, data processing questions, or anything this page does not cover, email info@ozwell.ai and we will get you a real answer.