Chatted 4 minutes — sounded well, mentioned her knee is stiff.
We are not sizing this on the US 65+ population. We don't own a network there, and the moat wouldn't travel. Canada first, properly. The US is a second, different conversation.
Iris says good morning, every day — the ritual she comes to love. It's the whole reason she picks up, and if she doesn't, somebody hears about it.
Today Iris screens what she hears on her calls for scam patterns and tells the family. The roadmap moves it into the network, before her phone rings — where an app can't follow.
No answer, no interaction — Iris calls. Still nothing, and it escalates on the family's standing order — and to a licensed monitoring partner as we sign one. Alongside a medical alert, never instead.
The family dashboard that fills itself. Every signal above feeds it live, so nobody has to update anything.
Papa proved health plans will pay for this — and that you cannot deliver it with gig workers. Iris costs cents a call and cannot assault anyone. That is why this works now and didn't in 2023.
“May attempt.” Never “will.” We do not promise to detect an emergency, and we do not promise a call — to the family or through the monitoring partner — will be placed, connected or answered. The liability in this category lives in the promise, not the capability — so we hold the capability and promise nothing. And the loudest line of all: “Evergreen works alongside a medical alert pendant, never instead of one. Keep the pendant.”
The daughter buys it; the mother consents to it. Under the TCPA an AI voice is an "artificial voice," and consent must come from the person being called.
Inference gets cheaper every quarter; we cascade routine turns to a small model, cache hard, cap the tail. Owning the network is worth about $2 a month — worth having, not the business. The business is charging a real price for real safety, while free AI companions race to zero and can never answer the 3 a.m. question.
A two-week legal-architecture sprint: the consent flow, the escalation ladder, the licensed-partner crisis protocol, and a full pass over our marketing copy with a litigator holding the pen. In this category your promise is your liability — Life Alert lost seven figures on its advertising, not its protocol. Cheap now, ruinous to retrofit onto 100,000 subscribers later.
The daily call is why she picks up; six months of Iris knowing what "a good morning" sounds like for Margaret is the thing a free app can never copy. The scam screening and the licensed-partner safety net are what a family pays for — loneliness is the need, and the relationship is the moat.