Who answers the phone, the resident or the AI?
The AI calls the resident on a schedule the family configured during onboarding. The resident answers their existing room phone or mobile. There is no app to download, no tablet to learn, and no new device on the nightstand. The agent introduces itself by name, identifies itself as an AI companion calling on behalf of the facility, and proceeds with the planned check in conversation. If the resident asks to speak with a human, the agent transfers to the on duty staff line. Calls happen at the cadence the family selected, ranging from daily to twice a week, with morning, afternoon, or evening windows.
How is this HIPAA compliant when the agent is processing voice from residents?
Every downstream provider that touches resident voice or transcript data is covered by a Business Associate Agreement. That includes the speech to text vendor, the LLM provider, the storage layer, and the family dashboard hosting. Recordings are encrypted at rest and in transit. Access is role based and audited. The pilot architecture maps every data flow against the HIPAA Security Rule and the relevant California elder care regulations. The facility owns the audit trail. We do not touch PHI in our offices, and the operator runs in a tenant the facility controls.
What does the agent actually do during a call?
It holds a light conversation. It asks how the resident is feeling, whether they ate breakfast, whether they slept well, whether they want a ride to the Tuesday class. It listens for wellness flags such as confusion, pain, low mood, or refusal of meals, and it routes those flags to the on duty nurse with the timestamp and the relevant transcript snippet. It schedules transportation to community classes when the resident says yes. It reminds the resident about a prescription refill that pharmacy noted is due. It does not diagnose. It does not give medical advice. It surfaces signal to humans who can act.
What does the family see?
A weekly summary email and a simple dashboard. The summary covers how often the resident answered the call, the topics they engaged with, the activities they accepted, the mood signal trend across the week, and any flagged items that staff already addressed. The dashboard has the same view plus the option to adjust the call schedule, change the topics the agent should focus on, and write a short note that the agent can mention on the next call: "your granddaughter started kindergarten this week." The family does not see raw transcripts unless the facility's policy allows it.
Does this replace the activity coordinator or the nursing staff?
No. It is the layer underneath them. The activity coordinator at most facilities is the bottleneck on family contact, transportation scheduling, and one to one engagement, because there is one of them and there are 80 to 200 residents. The Resident Companion handles the volume of routine touchpoints so the coordinator can spend time on the residents who actually need a person, the families who need a real call, and the activities that require a human to plan. Nursing staff get cleaner wellness signal earlier, so the escalation queue is shorter and better triaged.
What does it cost and how long to deploy?
Investment range: 30,000 to 80,000 dollars to deploy, plus 2,000 dollars per month per facility for the operating layer. Timeline: 6 to 10 weeks from contract to first resident calls. The variation in deployment cost depends on the existing PMS, the EMR integration, the transportation scheduling system, and the family communication channel the facility already uses. The 2,000 dollar per month line covers BAA hosted compute, telephony, transcript storage with retention controls, and the family dashboard.