AI at SeamlessMD

Yes, in specific, bounded ways — and SeamlessMD is deliberate about distinguishing AI from deterministic logic. The core clinical decision logic in the platform is not AI: alert thresholds, branching survey logic and risk stratification scoring are rule-based, fully transparent, configurable by the health system’s clinical leads and version-controlled.

AI is applied in four places: Seamless Answers (patient-facing conversational AI), natural language processing for free-text patient feedback, AI-assisted content translation, and analytics support for care teams. All AI features are optional and can be disabled.

Seamless Answers is SeamlessMD’s conversational AI feature for patients. It lets a patient ask a question about their care in their own words — “can I shower with my incision,” “when do I restart my blood thinner,” “is this much swelling normal” — and get an immediate answer at any hour. Answers are generated exclusively from the education content the patient’s own care team has reviewed and approved. It is live at numerous health systems such as North York General Hospital, Kingston Health Sciences Centre, Southlake Health, Sault Area Hospital and North Bay Regional Health Centre.

Seamless Answers uses a retrieval-augmented generation (RAG) architecture grounded strictly in the health system’s approved content library. The model retrieves only from that library at inference time and cannot access external sources or answer from unconstrained model knowledge. Questions falling outside the approved content are not answered speculatively — the patient is directed to contact their care team. This constraint is deliberate: it is what prevents the model from generating unsanctioned clinical guidance, and it is why a patient-facing generative AI feature can be deployed safely in a clinical setting.

Yes — the care team owns the approved content library that Seamless Answers draws from, entirely. If content is not approved by the health system, Seamless Answers will not use it. Health systems retain the same review, edit and update workflow over AI-answerable content that they have over the rest of their care plan content, and all content libraries are version-controlled with full change history.

Yes. Natural language processing is applied selectively to free-text patient responses — for example, sentiment analysis on patient experience survey comments to surface themes and flag concerns for quality improvement review. SeamlessMD also applies analytics across patient-reported data at the program level so care teams can see patterns without reading every individual response.