Skip to content

About

For everyone who has someone too far away to check in.

Millions of older people live alone while the people who love them live in another city, another country, another time zone. The phone calls that should happen don't, because everyone is busy and nobody wants to be a burden. Zika Care is that call.

Why a phone call

Because it’s the one piece of technology she already trusts. Every app, tablet, pendant and smart speaker asks an older person to learn something new, charge something, wear something or look at something. A phone just rings.

For someone with low vision, a phone is strictly better than a screen. For someone with a tremor, it’s better than a touchscreen. For someone proud, it’s better than a gadget that announces they need looking after.

Why conversation and errands, together

Conversation is what makes someone keep answering. Getting things done is what makes it matter. A companion that only chats is pleasant and easy to drop. A service that only runs errands is useful and cold. The auto-rickshaw that turns up at five is what turns a machine that checks on her into something that works for her.

Why we’re slow on purpose

We’re opening one country and one language at a time, and only once the voice has been tested on real phone lines with older people who speak that language, the crisis resources are mapped, and every provider in the path has agreed in writing not to train on her voice. That’s slower than launching everywhere. It’s the only way we know to be trusted by someone who didn’t ask for us.

The test we hold ourselves to

“This makes me talk to my mother more, not less.”

If a family can’t say that after a month, we haven’t done our job.

Where we are

Pre-launch. Taking waitlist sign-ups; not yet serving families.

Before serving anyone at scale, we’ll run the service by hand with a small group of families, to learn what matters before we automate it. Waitlist members in the first languages will hear from us first.

What we build on

Serious infrastructure, kept honest.

A real-time voice model

Conversation runs on a leading real-time speech model. Our own turn-taking sits on top, because the defaults were built for people who talk fast.

Rules outside the model

The things it must never do — ask for a PIN, name a condition, speak as a relative — are enforced by our own code, not left to a prompt.

No training on her voice

We only work with providers who contractually exclude training on her audio and words. Where that isn't possible, we don't offer the service.

Encrypted, logged, limited

Encrypted in transit and at rest, with a second layer on audio and transcripts. Every human access is logged. Only what a given moment needs is ever sent to a model.

Too far away? Not from their phone.