AI Integration with Eyefinity Virtual Visits

AI connects to three primary surfaces within Eyefinity Virtual Visits: the patient portal/messaging APIs for pre-visit intake and post-visit follow-up, the video platform integration layer (often via Zoom or similar partners) for in-visit transcription and real-time data capture, and the core scheduling and EHR modules to update visit records, trigger billing, and sync clinical summaries. The integration acts as a middleware layer that listens for events like a scheduled virtual visit, a launched video session, or a completed encounter, then executes AI-driven workflows.

A practical implementation wires a secure service to the PatientCommunication and Appointment APIs. For example, 24 hours before a visit, an AI agent can send a personalized intake message via the portal, analyze the patient's responses for urgency or needed prep, and flag the visit record. During the visit, a real-time transcription service, connected via the video platform's webhook, can provide a live transcript to the provider and highlight action items. Post-visit, another agent can draft a visit summary using the transcript and structured EHR data, submit it for provider review via an API, and then trigger automated follow-up messages or billing code suggestions based on the finalized notes.

Rollout should be phased, starting with non-clinical workflows like post-visit summary generation to demonstrate value without immediate clinical risk. Governance is critical: all AI-generated content must be clearly marked as a draft, require provider review and sign-off within the Eyefinity interface before becoming part of the legal record, and maintain a full audit trail linking the AI action to the user who approved it. This approach allows practices to capture efficiency gains in documentation and patient communication while maintaining compliance and clinical oversight.

The core integration connects to two primary Eyefinity surfaces: the Virtual Visits API for real-time session data (participants, duration, status) and the Practice Management API for patient records, schedules, and clinical data. A middleware layer—often deployed as a secure cloud service—acts as an orchestration hub. It listens for webhook events from Eyefinity (e.g., visit.scheduled, visit.started, visit.ended) and triggers corresponding AI workflows. For pre-visit triage, the system calls the patient record via the Practice Management API to populate a triage bot with context, then uses a tool-calling LLM agent to conduct a structured SMS or web-based interview. The agent's findings are written back to the patient chart as a structured note via the API, flagging urgency for staff review.

During the visit, the architecture supports real-time transcription by securely streaming audio from the Virtual Visits session (with patient consent) to a speech-to-text service. The transcript is processed in near-real-time by an LLM to identify key clinical points, action items, and questions. This enriches the provider's view within the telehealth interface via a side-panel or overlay, powered by a frontend component that fetches processed insights from the orchestration layer. Post-visit, the same transcript and any in-session annotations are used to automatically generate a SOAP note draft and a patient-friendly summary. The draft is inserted into the appropriate section of the patient's EHR chart via the Practice Management API, while the summary is queued for delivery through Eyefinity's patient communication channels (portal, SMS, email) following configurable staff review and approval steps.

Governance and rollout are critical. The implementation should use role-based access controls (RBAC) native to Eyefinity to ensure AI-generated content is only written by authorized service accounts. All AI interactions should be logged in a dedicated audit trail linking to the visit ID. A phased rollout typically starts with post-visit summarization in a single location, using a human-in-the-loop approval workflow where providers review and sign off on all AI-generated notes before they become part of the legal record. This builds trust and allows for prompt tuning. The architecture must also plan for handling PHI; all data in transit to and from AI models should be encrypted, and any external AI services must be covered under a BAA. The middleware should include circuit breakers and fallback logic to ensure visit workflows continue uninterrupted if AI services are degraded.

A production integration for Eyefinity Virtual Visits is built on a secure middleware layer that sits between the telehealth platform and AI services. This layer handles PHI de-identification before data is sent to LLMs for tasks like visit summarization, routes transcription audio through HIPAA-compliant speech-to-text services, and manages audit logs of all AI interactions. The integration connects to Eyefinity's APIs for the telehealth module, patient communication channels, and visit records, ensuring AI outputs are written back to the correct patient context and user session.

Rollout follows a phased, risk-managed path. Phase 1 often starts with a non-clinical pilot, such as an AI-powered pre-visit triage bot that handles administrative FAQs via the patient portal, building trust without touching clinical decisions. Phase 2 introduces in-visit ambient transcription for a single provider group, with outputs flagged as drafts requiring clinician review and sign-off within the Virtual Visits interface. Phase 3 expands to automated post-visit summary generation, where AI drafts follow-up instructions and visit notes that are routed through a configured approval workflow before being attached to the patient record or sent via Eyefinity's messaging system.

Governance is enforced through technical and policy controls. Role-based access controls (RBAC) within Eyefinity determine which staff can trigger AI features or approve AI-generated content. A human-in-the-loop requirement is baked into clinical workflows, ensuring a provider validates all AI-suggested clinical text. All AI interactions are logged to a secure audit trail, capturing the source data, prompt, model used, output, and approving user for compliance reviews. This structured approach allows practices to capture efficiency gains—turning documentation from a post-visit chore into a reviewed draft in minutes—while maintaining rigorous oversight over patient care and data.