Last Updated: April 16, 2026
Auxo Health Solutions ("Auxo") uses artificial intelligence ("AI") and machine learning ("ML") to support certain features of the Service. This notice describes how we use AI, what we don't do with AI, and how we help clinicians use AI outputs safely.
What AI Features We Offer
Current AI-assisted features include:
- Clinical documentation assist — drafting progress notes, treatment plans, assessments from structured and unstructured inputs
- Utilization review summaries — compiling medical necessity rationale for concurrent review and appeals
- Clinical question-answering (RAG) — answering clinician questions about a patient's record, always with source citations to the underlying data
- Risk flagging — identifying documentation gaps, expiring authorizations, AMA risk indicators
- Ambient scribing — generating draft clinical notes from audio recordings of clinician–patient conversations (with patient consent)
- Agentic billing — generating draft claims, denial responses, and appeals for human review
Additional AI-assisted features may be added; the public sub-processor list identifies the AI providers we use.
What Clinicians Must Do When Using AI Outputs
Every AI-generated output is a draft. A licensed clinician must independently review it before relying on it for any clinical purpose. This applies to clinical notes, medication suggestions, risk scores, treatment plans, assessments, discharge summaries, and any other content the Service produces through AI.
Specifically:
- Review sources: Each AI-generated recommendation cites its sources. Clinicians should verify the sources support the recommendation before relying on it.
- Correct errors: AI may misread records, hallucinate facts, or miss nuance. Edit, correct, or reject outputs as clinically appropriate.
- Attest to the final version: The clinical note, plan, or order that is signed is the clinician's work product, regardless of how it was drafted.
- No patient-facing AI advice: Patient-facing AI-generated content includes a notice that it was generated by AI and directs the patient to contact a licensed provider for clinical questions.
What We Don't Do with AI
- We do not train foundation AI models on Customer Data or PHI. Our AI sub-processors (including Anthropic, Claude; Deepgram; and any others listed on the sub-processor page) are contractually bound by zero-data-retention agreements for Auxo's account.
- We do not use AI to make autonomous clinical decisions. Every clinical decision requires a licensed clinician in the loop.
- We do not process data from AI features through vendors without BAAs. All AI sub-processors that see PHI have executed BAAs with Auxo.
- We do not claim the Service is a medical device. The Service is a decision-support platform; clinical decisions are made by clinicians.
Regulatory Posture
The Service is designed to operate within the Clinical Decision Support (CDS) exclusion from the FDA medical-device definition, under 21 U.S.C. § 360j(o)(1)(E) and the FDA's September 2022 final guidance. The Service:
- Does not acquire, process, or analyze a medical image or in-vitro-diagnostic signal
- Displays, analyzes, or summarizes medical information
- Provides recommendations to healthcare professionals (not directly to patients for autonomous use)
- Allows the healthcare professional to independently review the basis for recommendations — we cite sources
We also comply with:
- California AB 3030 — AI-generated communications directed at patients include the required AI-generated disclaimer.
- Texas HB 2060 — AI risk assessment and consumer notification obligations.
- HHS OCR AI guidance — documented risk assessment of AI features that process PHI.
Changes
We will update this notice when we add new AI features or change how we use AI.
Contact
For questions about our AI practices, email privacy@auxohealthsolutions.com.