The idea of an AI that answers calls in your voice is new enough that most families have a lot of questions before they're ready to consider it. This post is those questions, the real ones, asked honestly, with the straightest answers we can give.
Does it actually sound like me?
It depends on what "sound like you" means, and it deepens over time.
Day 1, after Session 1: You complete a short interview session with our setup specialist Sarah, about 7 minutes of natural conversation. An Instant voice presence is built from your audio alone, it captures your timbre, your accent, your rhythms. Recognizable as you immediately.
Sessions two and three, plus optional Advanced Training: After two more short interview sessions during onboarding, your model deepens into your Refined Voice, the point where most listeners on a phone call cannot reliably tell the difference. Whenever you want to sharpen it further, open the Voice tab in your dashboard and add another short session. Real conversations between the companion and your loved one are never recorded for training.
Refined Voice, deepened on your terms: The model can keep deepening with optional Advanced Training sessions. At this point most listeners on a phone call cannot tell the difference. It sounds like you actually called.
For people with dementia, who respond primarily to emotional familiarity rather than technical perfection, the voice tends to land as recognizable and comforting from the beginning. The part of the brain that recognizes a beloved voice is not doing critical analysis. It's doing pattern recognition, and the pattern is there from day one.
What if they figures out it isn't really me?
Most people living with dementia don't question the calls. They feel the comfort and warmth of a familiar voice, and that's what they respond to.
If your loved one does express uncertainty, "is this really you?", KindredMind is trained to respond gently and reassuringly, without arguing or insisting. Something like: "Of course it's me. I'm always here when you call." The goal is never to assert facts over their experience, but to provide the emotional reassurance they's looking for.
In practice, the moments that trigger uncertainty tend to be specific situations, a particular topic they's sensitive about, a voice quality that shifts during a certain emotional register. If you notice patterns through the call summaries, you can add knowledge to the system that helps it navigate those moments better.
How long does setup take?
The core setup, recording your voice, building the basic knowledge base, connecting the phone number, takes most families about 30 minutes across the onboarding process.
The voice is ready to use immediately after your recording session. The knowledge base can be added to over time, it's designed to grow as you think of things, not to require comprehensive input before it starts working.
What does it actually know about them?
Whatever you tell it. That's the honest answer.
During onboarding, you provide information about: their daily routine, their care setting, staff names and friends, their favorite topics and memories, what upsets them and what calms them, the questions they always ask and the answers they need to hear, upcoming plans, family news, and anything else that makes them who they are.
The more you give it, the more every conversation feels specific to them rather than generic. A conversation that mentions their nurse's name, the bingo game this afternoon, the fact that Jake had a hockey tournament last weekend, that feels like it could only be about them. That specificity is what makes the familiar voice feel genuinely present.
What happens in a real emergency?
If your loved one indicates they is in pain, is frightened of something specific, is describing a physical problem, or says anything that sounds like a genuine emergency, KindredMind alerts you immediately. You receive a notification and can join the call instantly with one tap.
The AI is also trained to encourage their to use their call button or speak to staff if they is at a care facility, and to tell their you are on your way. The goal is always to get real human help in situations that require it, KindredMind does not attempt to manage medical situations.
You are always in the room. KindredMind gives you the ability to monitor and join, and never takes that ability away.
Is my family's information safe?
The privacy of your family's most intimate conversations is something we take seriously, because we are a family that has lived what you're living.
- Your voice recordings are used to build your model and then automatically deleted within 24 hours of your model being ready.
- The voice presence is built only from your own short interview sessions with our setup specialist, never from real conversations with your loved one.
- Call transcripts are automatically deleted after 90 days.
- Your family's private conversations never train our models.
- Built in Canada and designed to meet both Canadian PIPEDA and US privacy standards.
The full accounting of our privacy and safety practices is at How We Protect Them.
Is this right for our family?
The families who find KindredMind most valuable tend to share a few characteristics, and the families for whom it may not be necessary right now have different ones.
Built for you if:
High-volume inbound repetitive calling driven by separation anxiety
Calls occurring at all hours, including overnight
Caregiver experiencing meaningful burnout from call volume
Mid-to-late stage, calls are emotionally driven rather than content-specific
May be more than you need if:
Earlier stage, they can still follow a schedule
Calls are fewer and more coherent
A structured check-in routine might address the need
If the call volume has reached a point where it is affecting your sleep, your work, and your mental health, if you have reached or are approaching the impossible moment of considering taking the phone away, then yes. This was built for you.
KindredMind is set up and authorized entirely by the caregiver who holds legal decision-making authority for the person with dementia — typically a spouse, adult child, or the holder of a Power of Attorney for Personal Care. This is the same framework that governs every care decision made on behalf of a person who lives with dementia. For a full treatment of the ethics and consent framework, read our ethics position.
Ready to see how it works? Setup takes about 30 minutes, and the voice is ready immediately.
How Families Use It Day to Day
The questions above address the concerns families have before they start. What families most often want to know after they have been using KindredMind for a few weeks is: how does this actually fit into real life? Here is what that looks like in practice.
The overnight calls
For most families, the most immediately transformative change is the 2am call. Before KindredMind, a nighttime call meant a choice: answer and be awake for an hour trying to settle them, or let it go and lie awake with guilt. After, the call is answered. In your voice. With warmth and the right information about where they is and who is with them. They settle. You sleep. The call happened, they were cared for, and you were not the one who had to carry it alone.
Families consistently report that recovering their sleep is one of the most significant effects. Not just practically, but emotionally. The background vigilance, the constant awareness that the phone might ring at any moment and you might have to perform the emotional labour of a fully present call, is exhausting in ways that go beyond the actual calls themselves. Having a system that handles those calls changes the quality of quiet time.
The work-hour calls
The second most common situation: the call that comes at exactly the wrong moment. In a meeting you cannot step out of. In the middle of something with a real consequence if interrupted. Before: you watch the phone ring, feel the guilt, try to concentrate, and call back when you can. After: KindredMind answers. You get a summary. They was worried about whether dinner was at the usual time, it is, and they settled within a few minutes. You finish your meeting.
The cumulative effect of this is hard to describe until you have lived it. It is not just the individual calls that are handled. It is the removal of the constant background calculation, which calls can I afford to miss, which can I afford to answer, what happens if I get the math wrong. When the answer is always "the call will be answered well," that calculation stops. Caregiver burnout is what happens when that calculation never resolves.
Building the knowledge base over time
The system gets better as you add to it. Most families start with the basics, daily routine, care staff names, key reassurances, and then add to the knowledge base as they notice patterns. They always asks about their sister on Tuesday afternoons. They worries about whether the cat has been fed. They needs reassurance about a specific upcoming appointment. Adding these specific details means the conversations feel more and more particular to them, not a generic AI response, but a response that could only be about their life.
The call summaries accelerate this. Every call produces a summary of what was discussed and what seemed to settle them. Over time, those summaries show you patterns you might not have noticed: what topics they returns to consistently, what time of day the calls cluster, which reassurances land and which do not. That information makes both KindredMind interactions and your own calls with their better. More on how the knowledge base and call handling work is on the product page, and on how we protect the privacy of your family's conversations.