Clinical Intelligence

How Welmivia uses AI
and how we don't

Patients increasingly ask this, and they deserve a direct answer. Here is exactly how AI is and is not used in your care.

Two systems, two purposes

We use technology in two distinct ways, and the distinction matters enough to explain rather than blur.

Diagnostic assessment tools

Tools like our ADHD assessment engine were built by Dr. Sonia Mia Diaz, M.D., using structured clinical criteria and standard validated scales. These run on local, on-premise hardware, not the cloud, and assessment data is not stored on the device after your results are recorded in your chart. That significantly reduces the exposure if a device were ever compromised: there is no local repository of patient data sitting on a hard drive waiting to be found.

Documentation support

We use AI to help draft your Welmivia Blueprint after a visit. Every framework, treatment pathway, and recommendation logic it draws on was written by Dr. Diaz, not generated by AI. The knowledge base is closed: a structured application of clinical protocols she has already built, based on current evidence and clinical guidelines, not an open model improvising on your case. It runs on HIPAA-compliant cloud infrastructure under a signed Business Associate Agreement. Every Blueprint is personally reviewed, edited, and finalized by Dr. Diaz before it reaches you.

What never happens

  • Your data is never used to train an AI model, ours or anyone else's.
  • No protected health information passes through consumer AI tools. Products without a signed Business Associate Agreement, including Gemini in Chrome and general ChatGPT, are explicitly excluded from any workflow that touches your chart.
  • AI does not diagnose you, does not make treatment decisions, and does not originate clinical recommendations. It drafts using guidance that a psychiatry-trained physician has already written, and she reviews, edits, and finalizes every output before it reaches you.

Why we built it this way

Most psychiatric practices are quietly adopting AI tools without telling patients how they work, and some are letting general-purpose AI generate clinical content directly. We do neither. Our knowledge base is proprietary and physician-authored; AI's job is administrative, not clinical. Precision psychiatry only works if the underlying process is trustworthy, and trust requires specifics, not reassurance.

The bigger picture on AI in psychiatric care

AI is moving into medicine faster than most patients realize, and psychiatry is no exception. Used well, it can reduce the administrative burden that pulls physicians away from patients, catch details a rushed note might miss, and free up visit time for the actual clinical conversation. Used carelessly, it can quietly insert generic, non-individualized content into a patient's chart, or route sensitive mental health information through tools that were never built to protect it.

The difference is not the technology. It is who controls it, what it is allowed to touch, and whether a physician remains the final decision-maker. That is the standard we hold ourselves to, and the standard we think patients should expect from any practice using AI in their care.

A deliberate choice

The decision to incorporate AI into Welmivia was not made lightly. Physicians at large institutions are often handed AI tools they didn't choose and can't fully see inside, with little say in how those tools work or what infrastructure they run on. As an independent practice, Dr. Diaz built and chose these tools herself, deliberately and selectively, rather than adopting a broad institutional system by default. That includes weighing the resource cost of running AI infrastructure at all, and using it narrowly and intentionally rather than layering on tools for their own sake.

That deliberateness is also what makes it useful to patients. It supports faster turnaround on evaluations and Blueprints, shorter wait times for appointments, and a more consistent standard of care across every patient, without giving up the physician judgment that makes the care personal.

Visit recording: Doximity Scribe

During some visits, we may use Doximity Scribe, a HIPAA-compliant tool built for clinicians, set to "Dictation only" mode, strictly to record and transcribe your visit. Dr. Diaz may run that transcript, alongside her own clinical notes, through the same physician-built Welmivia Clinical AI system described above, as a check: it validates her clinical plan against everything shared in the visit, ensuring nothing gets missed.

  • Doximity's Business Associate Agreement is established automatically at registration and explicitly covers Scribe as one of its designated secure tools.
  • Audio from your visit is not stored. It is processed in real time to produce a transcript and then immediately discarded. Because nothing is retained, there is no recording that could ever be used to train an AI model.
  • This check never changes your chart or your Blueprint on its own. Dr. Diaz reviews the result and decides what, if anything, it changes.
  • The transcript is retained for 29 days, giving time for review and chart integration, then automatically deleted.

If you would prefer a visit not be recorded, tell us and we will document your visit the traditional way instead.

Questions about how a specific tool handles your information? Ask at your next visit, or reach out directly. We would rather over-explain this than have you wonder.

Book Your Evaluation
admin@welmivia.com  ·  (434) 623-0003

Get in touch

Contact Welmivia

Questions about services, scheduling, or what to expect? We respond within one business day.

Existing patients: log into the patient portal for clinical messages.

Message received

Thank you for reaching out. We'll be in touch within one business day.

For general inquiries only. Not for urgent medical concerns.