QLC Digital health

The gap Eli sits inside

A particular kind of loneliness arrives with a new baby. The house fills with visitors and casseroles and advice, and somewhere inside all of that, a woman is sitting up at three in the morning wondering why she doesn't feel the way everyone keeps telling her she should. She might be exhausted. She might be sad. She might be terrified. She almost certainly isn't going to phone a clinic about it in the morning.

That's the gap Eli was designed to fix.

Eli is a chatbot. The word doesn't really do it justice, so set it aside for a second. Eli is a presence on WhatsApp, the same app a woman already uses to send pictures of her baby to her mother, that introduces itself like a friend, asks her name, and gently starts a conversation about how she's really doing. It's built for the perinatal period: the months of pregnancy and the year or so after birth, when mental health risks are highest and the support around women is often weakest.

It isn't a therapist, and it isn't pretending to be one. It's a careful, kind first step toward one.

Why a chat, and why WhatsApp

The hardest part of maternal mental health is almost never the diagnosis or the treatment. It's the first sentence. Saying "I am not okay" out loud, to another human, in a room, is more than many women can manage when they're already running on two hours of sleep and a thick layer of guilt. Apps with names like Depression Helper sit unopened on phones because nobody wants the icon visible when their partner picks up the screen.

WhatsApp removes all of that. Nothing to download. No new account to create. No icon to hide. A woman messages a number the way she'd message a friend, and a friend writes back.

That isn't a small design decision. It's the whole point. Every other choice in Eli follows from it.

What a conversation feels like

Eli opens with its name and a sentence that sounds like a person wrote it, because a person did. It asks what the woman wants to be called, then uses that name throughout. It promises privacy in plain language and links to a real privacy policy, not a wall of legalese.

Then it asks where she is in her parenting journey, and what's weighing on her today. The options are written in the language women actually use: I feel sad or disconnected from my baby. My baby's in the NICU. I'm stressed and anxious. For each, Eli responds with a few sentences of warm, validating acknowledgement, the kind a thoughtful friend might offer, and then gently moves toward either a brief grounding exercise or a screening.

The exercises are deliberately small. For anxiety, Eli walks her through a thirty-second sensory grounding practice: five things you can see, four things you can feel, three things you can hear, two things you can smell, one good thing about yourself. For sadness, it offers a single phrase to repeat with a hand over the heart: This is a hard moment. All mothers have hard moments. May I be kind to myself right now. These aren't cures. They're anchors, offered without pressure. The woman is told, every step of the way, that she can stop.

The careful part

When the moment's right, Eli offers a screening: ten short questions known as the Edinburgh Postnatal Depression Scale. The EPDS is a tool midwives and clinicians around the world have used for decades to surface signs of postnatal depression. Eli asks the questions one at a time, conversationally, with the option to pause built into every step.

Eli scores the answers privately and uses the score to decide what to offer next. A low score ends with a kind goodbye and an invitation to a community group of other mothers. A middle score invites her to consider a conversation with a professional, or with a peer group. A higher score leads to a clearer recommendation, with a member of the care team reaching out within seventy-two hours.

The tenth question of the EPDS asks about thoughts of self-harm. If a woman answers in a way that signals risk, Eli stops everything. It thanks her for her honesty, because that's the right thing to do, and because honesty in that moment is the bravest thing a person can offer. Then it shows her a twenty-four-hour helpline number. The conversation doesn't move forward until she has confirmed she's seen it.

There's also a single word, SOS, that a woman can type at any point from the very first message to the very last, and it will pull her straight into a crisis check-in. Eli stop ends the screening. Eli back steps her one question back if she wants to change her answer. Eli restart begins again. These commands exist because a woman in distress shouldn't have to think hard about how to use the tool that's trying to help her.

For the organisations watching this

If you run a clinic, a maternal health programme, a community group, or a faith organisation, and you've been wondering whether something like Eli could work for the people you serve, here are a few things worth knowing.

The script (every question, every option, every branch) lives in a single human-readable file that your clinical or programme staff can read and edit. Adding a country, softening a sentence, changing a referral pathway: none of that requires an engineer. The technical team sets up the rails. Your team writes the conversation.

Safety is built into the script, not bolted on. The crisis routing, the helpline number, the SOS command: these aren't features that got added later because someone asked. They were part of the design from the first message.

Privacy is a stance, not a checkbox. Conversation state lives in an encrypted store that automatically expires after forty-eight hours, so a woman's words don't sit on a server longer than they need to. Only what the care team needs to follow up on is kept for the long term.

The system is built on the kind of plumbing you'd want under something this sensitive. It runs on Elixir, a programming language built for handling many conversations at once without falling over. That's the same family of technology that powers WhatsApp itself. It connects through Meta's official WhatsApp Business API. The screening is a recognised clinical instrument, not something invented for the demo. The care team works in tools they already know.

And the pattern travels. The same shape, a warm scripted conversation on WhatsApp with a safety net underneath and a human team behind it, could be adapted for adolescent mental health, for men reluctant to walk into a clinic, for post-surgical follow-up, for chronic disease check-ins, for any context where the first sentence is the hardest one.

The quiet point

There's no magic in Eli. It doesn't improvise. It doesn't diagnose. It doesn't pretend to understand more than it does. What it does is sit in a woman's phone, in the app she already trusts, and lower the threshold for asking for help from I would have to walk into a clinic and explain everything from the beginning to I would have to send a text.

For a lot of women, that's the difference between getting help and not. That's the whole project.