RPM Manual
The practical 2026 guide to device rules, day thresholds, management time, and audit defensibility for Remote Patient Monitoring.
Read the RPM Guide →Amy: How AI Prioritizes Clinical Workflow & Fixes Device Issues
The Bottom Line
- Active Triage: Amy doesn't just "chat"; she analyzes patient data against your care plan to suggest specific next steps (e.g., "Send device guide").
- Knowledge Retrieval (SemDB): The AI accesses your internal PDF library (like device manuals) to answer specific questions without hallucinating.
- Action, Not Just Text: You can execute clinical actions--like sending a text or logging a call--directly from the chat interface.
- Hallucination-Free: By using an ontology engine (Buffaly), the system ensures answers come from your approved medical rules, not the open internet.
How We Prevent "AI Hallucinations" in Clinical Care
A common fear with AI in healthcare is "hallucination"--when a model invents facts. As shown in the video, Amy is different because she is not a generic chatbot. She is an Agentic Retrieval System anchored by two specific technologies described below.
1. SemDB (Semantic Database)
Instead of guessing, Amy looks up answers in a "walled garden" of your specific documents (e.g., the Omron BP Monitor Manual or your clinic's hypoglycemia protocol). She can only provide answers found in these approved texts.
2. Buffaly (Ontology Engine)
Buffaly acts as a security layer between your patient data and the Large Language Model (LLM). It ensures that clinical reasoning follows a strict logical structure (an "ontology") so the AI understands that High BP + Headache = Urgency.
The "Human-in-the-Loop" Workflow
Notice in the video that Amy does not auto-send the medical advice. She proposes the action ("Should I send the replacement guide?"), and the human clinician clicks "Yes." This keeps the license-holder in control while removing 90% of the manual research time.
Video Transcript
0:00 [Justin] Hi, I'm Justin Brochetti from FairPath. In the previous video, we built a care plan for a patient with hypertension and diabetes. Now, we have data flowing in, but we also have a patient who might be having technical trouble. Instead of digging through PDFs or calling tech support, I'm going to ask Amy for help.
0:20 I'm inside the patient's workspace. I can open the Amy chat window on the right. I'm going to type: "The patient is reporting that their blood pressure cuff feels loose and isn't giving a reading. What should I do?"
0:35 Now, watch what happens. Amy isn't just making up an answer. She is querying SemDB--our internal knowledge base where we have stored the user manuals for all our devices. She analyzes the query and retrieves the specific section on "Cuff Fitting and Replacement" for this specific model.
0:52 [Amy Output] "Based on the device assigned to this patient (BodyTrace BT-105), a loose cuff often indicates incorrect sizing or placement. I have located the 'How to Replace and Fit Your Cuff' guide in the documentation. Would you like me to send this guide to the patient via SMS?"
1:10 [Justin] This is the power of agentic AI. She didn't just tell me how to fix it; she prepared the solution for the patient. I can simply click "Send SMS" right here in the chat.
1:22 The system automatically formats the text message, attaches the PDF guide from our library, and sends it to the patient's phone. If I look at the audit log, that interaction is timestamped and recorded. I didn't have to leave the screen, search Google, or find a file. Amy turned a 15-minute support ticket into a 30-second interaction.
1:40 Next, we'll look at how to review the clinical data that starts coming in once the device is working. But for now, that is how you use Amy to unblock patient issues instantly.
Stop Being Tech Support. Start Being Clinicians.
See how FairPath's AI agents handle the noise of device logistics and data filtering so your team can focus on care.
