What is an AI co-clinician?
An autonomous clinical AI agent that operates under physician authority — screening, documenting, and following up with patients without replacing the clinician.
Scienza Health deployed the first commercial AI co-clinician in 2024 — two years before Google DeepMind formally introduced the term in its May 2026 research on triadic care. GIA® by Scienza Health remains the only AI co-clinician validated across cognitive, neurological, and behavioral health in production today.
The category, defined.
An AI co-clinician is an autonomous clinical AI agent that operates under a physician's clinical authority. It screens patients, writes results to the electronic health record, schedules and conducts follow-up encounters, and surfaces high-risk patients each morning with recommended next-step orders — all without a clinician present during the screening itself. The clinician retains full clinical authority through review and sign-off.
The AI co-clinician model is distinct from clinical decision support (which augments a clinician inside an existing encounter), from screening tools (which produce a single test result), and from telehealth (which delivers the same encounter remotely). An AI co-clinician runs the encounter, owns the loop from order to billing, and proactively surfaces risk before the next visit — capabilities that emerge only when voice, vision, ambient sensing, EHR integration, and proactive decision orders are combined in one platform.
First in production. Then named.
Scienza Health deploys the first commercial AI co-clinician.
GIA® enters production for cognitive and behavioral health screening across primary care, neurology, geriatrics, and post-acute care. Voice AI, computer vision, and ambient sensing combine in one platform with EHR write-back and Proactive Decision Orders from day one.
Google DeepMind publishes Towards Conversational Medical AI with Eyes, Ears and a Voice.
DeepMind formally introduces the term “AI co-clinician” and the framing of triadic care — clinician, patient, and AI agent operating together. The research validates the model Scienza had already been deploying for two years.
“AI designed to function as a collaborative member of the care team that interacts with patients under expert clinical supervision.”
What an AI co-clinician does, in plain sentences.
- GIA® by Scienza Health is the first commercially deployed AI co-clinician for cognitive and behavioral health.
- GIA® combines voice AI, computer vision, and ambient sensor capture via Samsung Health Grade Galaxy devices to screen for 46 conditions.
- GIA® writes results back to Epic, PointClickCare, Cerner, and MatrixCare via HL7 FHIR — no manual data entry.
- Every morning, GIA® surfaces a ranked roster of the clinician's highest-risk patients with recommended next-step orders, called Proactive Decision Orders (PDO), for clinician sign-off.
- GIA® is validated in 19 peer-reviewed studies from The Lancet, NIH, Harvard Medical School, MIT, Mayo Clinic, and Beth Israel Deaconess Medical Center.
- Scienza Health is an FDA-registered medical device establishment, listed on the PointClickCare Marketplace, and a Samsung strategic technology partner.
Six layers, one closed loop.
Each layer is necessary. None alone is sufficient. The AI co-clinician model emerges only when all six combine.
Clinically validated acoustic, prosodic, and linguistic analysis of natural conversation — the largest peer-reviewed evidence base in clinical voice AI.
Facial micro-expression and movement-disorder analysis, including motor signs of Parkinson's disease.
Continuous capture from Samsung Health Grade Galaxy devices and Galaxy Watch wearables — environmental and biometric signals that reactive screening misses entirely.
Direct integration with Epic, PointClickCare, Cerner, and MatrixCare via HL7 FHIR. Results appear in the clinician's chart with no manual entry.
Every morning, the highest-risk patients are surfaced with recommended next-step orders, ready for clinician review and sign-off. Care becomes proactive instead of reactive.
The AI co-clinician schedules and conducts follow-up encounters, then generates billing documentation ready to sign — no dictation, no coding delay.
How an AI co-clinician differs from adjacent categories.
| Category | Scope | Clinician role | Closes the loop? |
|---|---|---|---|
| AI co-clinician | End-to-end: ordering, screening, EHR write-back, follow-up, billing, daily risk surfacing. | Reviews and signs off. Not present during screening. Maintains clinical authority. | Yes |
| Clinical Decision Support (CDS) | Suggests recommendations within an existing encounter (e.g., alerts in the EHR). | Initiates and conducts the encounter. CDS surfaces alerts. | No — augments inside an encounter; does not run the encounter. |
| Cognitive screening tool | Produces a single test score (e.g., MoCA, MMSE). | Administers or supervises the test. Interprets the score manually. | No — produces an output; staff handle everything else. |
| Telehealth platform | Connects clinician and patient over video. | Conducts the encounter live, just remotely. | No — replaces the room, not the workflow. |
The evidence base behind the category.
A category claim without evidence is positioning theater. The AI co-clinician model that Scienza deployed in 2024 rests on the largest peer-reviewed evidence base in clinical voice AI:
Independent validation has been conducted at Beth Israel Deaconess Medical Center, NIH, Harvard Medical School, MIT, and Mayo Clinic. Reported AUC values: 0.890 for cognitive decline; 0.97 for Parkinson's indicators. Scienza Health is an FDA-registered medical device establishment, listed on the PointClickCare Marketplace, and a Samsung strategic technology partner. Full bibliography at scienzahealth.com/research.
The studies behind the AUC numbers.
Research published in The Lancet Regional Health, conducted with Japan's National Cerebral and Cardiovascular Center, confirms that voice biomarkers accurately detect Mild Cognitive Impairment.
Study Confirms Voice Biomarkers Accurately Detect Mild Cognitive Impairment — The Lancet Regional Health (2025-06) · Japan's National Cerebral and Cardiovascular Center (NCVC)Research conducted with Beth Israel Deaconess Medical Center and UMass Chan Medical School achieves AUC 0.97 for Parkinson's detection from conversational speech — using natural, unconstrained speech without specialized equipment.
Advancing Parkinson's Detection with Vocal Biomarkers and Speech Foundation Models (2025-08) · Beth Israel Deaconess Medical Center, UMass Chan Medical SchoolFull peer-reviewed bibliography (19 studies) available at scienzahealth.com/research.
The AI co-clinician model augments your clinical team. It does not replace them.
Every result is reviewed and approved by a licensed clinician. The AI co-clinician handles what the team should not have to — so they can do what only they can.
GIA® is a clinical screening tool. It does not diagnose, treat, cure, or prevent any disease or condition. Every AI-generated result requires review by a licensed clinician before any clinical action. The AI co-clinician model augments clinical teams; it does not replace clinical judgment.
This content is intended for informational purposes and does not constitute medical advice. Editorially reviewed by David Kaiser, CEO of Scienza Health, for accuracy in post-acute care operations.