Demo
Prenatal care in SUS, Brazil's public health system, on track via WhatsApp
Lia answers questions with texts approved by the healthcare team and reminds women of appointments, tests and vaccinations. She does not advise on medicines, diagnose or decide on treatment. She recognises warning signs and directs pregnant women straight to a maternity hospital.
The conversation plays on its own. When buttons appear, tap one: your choice changes the path.
In testing, with no real pregnant patients. Lia is not yet for sale. The screens use fictional names, dates and places.
Liaassistente da equipe da UBS
Our goals for SUS
- Improve the quality of prenatal care
- Reduce the number of complications
- Reduce unnecessary interventions
- Reduce costs
21.452births in Tocantins in 2024SES-TO, preliminary Sinasc data, February 2025
73,7%of live births in Araguaína in 2024 had 7 or more prenatal visits, versus 78.0% in TocantinsSINASC 2024, via TabNet
Hear Lia reply
Questions written by the team, the way a pregnant woman writes, and Lia's answers in the voice she hears.
Lia speaks Portuguese and was tested in SUS, Brazil's public health system. On the phone, the messages stay in the original Portuguese, with the translation below.
Answers from version 1.1d, read unchanged by a synthetic voice. No message came from a real pregnant woman. The maternity hospital named is the reference one used in the tests.
Liaassistente da equipe da UBS
Use the arrow keys or swipe the phone to change conversations.
How it works
Safety rules are built into the system and can be tested.
Entry via the UBS QR code
In the project, a pregnant woman starts the chat through the QR code on a poster at the UBS, the local public health clinic.
Approved text
Artificial intelligence (AI) identifies the topic of the message and selects an answer approved by the healthcare team. It does not create health advice.
Alarm before the AI
The team approves the alert words. Lia directs pregnant women straight to a maternity hospital or 192, Brazil’s ambulance number, and AI cannot cancel this warning.
The team makes care decisions
Questions about medicines, diagnosis or treatment go to the UBS team. Healthcare professionals make care decisions.
Reminders throughout pregnancy
Lia reminds pregnant women of their appointments, tests and vaccinations. For an appointment, she sends a reminder the day before, on the morning of the appointment and 20 minutes beforehand, using the UBS schedule.
Answers reviewed, errors published
Healthcare professionals assess the answers without knowing who produced them. Errors found in testing are shown on this page.
Liaassistente da equipe da UBS
Ontem
Hoje
Lia, lembrete: Amanhã é a sua consulta de pré-natal na UBS, às 8h. Leve a sua Caderneta da Gestante. Se não puder ir, me avise que eu ajudo a remarcar.18:00
Lia, lembrete: Bom dia. Hoje é a sua consulta de pré-natal na UBS, às 8h.06:30
Lia, lembrete: A sua consulta começa em 20 minutos. Até já.07:40
Appointment reminderExample
Translation
Lia: Tomorrow is your prenatal appointment at the UBS, at 8 a.m. Bring your pregnancy booklet (Caderneta da Gestante). If you can't go, let me know and I'll help you reschedule.
Button: I confirm
Button: I need to reschedule
Button: I have a question
Lia: Good morning. Today is your prenatal appointment at the UBS, at 8 a.m.
Lia: Your appointment starts in 20 minutes. See you soon.
Responsible AI: what Lia does differently
Lia's safeguards, the stage she is at, and where she still makes mistakes.
| Feature | Lia | Regular AI chatbots | Call centers with agents | SMS messages |
|---|---|---|---|---|
| Does not make up health text: the answer comes from texts written by the team | LiaYes | Regular AI chatbotsNo | Call centers with agentsPartly | SMS messagesYes |
| Emergency alarm the AI cannot turn off | LiaYes | Regular AI chatbotsNo | Call centers with agentsPartly | SMS messagesNo |
| Mental health care, with CVV 188, Brazil's free emotional support line | LiaYes | Regular AI chatbotsPartly | Call centers with agentsPartly | SMS messagesNo |
| Tests without patient data | LiaYes | Regular AI chatbotsNo | Call centers with agentsNo | SMS messagesNo |
| Replies in text and audio on WhatsApp | LiaYes | Regular AI chatbotsPartly | Call centers with agentsPartly | SMS messagesNo |
| Tested by the health team, with the errors published | LiaYes | Regular AI chatbotsPartly | Call centers with agentsPartly | SMS messagesPartly |
- Area
- Prenatal care through SUS, from a positive pregnancy test to after birth. The project starts in Araguaína (TO) and is designed for all of Brazil.
- Stage
- In testing, using only messages made up by the team. Lia does not yet support real pregnant patients and is not for sale.
- Type of AI
- Identifies the topic of the message and selects an approved answer.
- Data
- No personal data. The tests use only made-up messages.
- Oversight
- The health team decides on care and reviews the conversations every 15 days.
- Evidence
- Version 1.0 bench test, with 60 messages written by the team, on October 30, 2026: 15 of 16 emergencies referred to the maternity hospital or to 192 (93.8%), 6 of 6 medicine requests passed to the team, and no alarm cancelled by the AI. The same action in 59 of 60 messages across the two rounds.
Published errors
Bench test of version 1.0, October 30, 2026
- 1 of 16emergencies did not receive emergency guidance. A woman who had just given birth reported troubling thoughts and was advised to call CVV, Brazil’s emotional support helpline.
- 6 of 39messages with no emergency triggered the alarm (15.4%), such as the baby moving little and extreme tiredness.
- 3 of 5distress messages reached the CVV 188 level. The other 2 got the emergency guidance.
- 40 of 60received the correct action (66.7%). In most errors, Lia passed the question to the team: this happened with 14 questions that already had approved answers.
The fixes go into version 1.1, with a new test on new messages. The conversations on this page come from 1.1d, which has not had that test yet.
Projects and partners
Lia's way of working fits other kinds of care in SUS: text approved by the team, urgent cases referred right away, and the clinical decision always with the health team.
Pré-natal em Dia
Pregnancy and after birth
This is the Lia on this page.
Mama em Dia
Breast cancer screening
The study plans to invite women aged 50 to 69 for mammograms and follow each woman through to treatment. It is planned for UBS clinics in Araguaína from 2027.
Crônicos em Dia
High blood pressure and diabetes
Lia would remind people with high blood pressure or diabetes of their appointments and tell the UBS team who has gone without follow-up.
Areas researched
A SULEIMAN study from October 2026 compared 14 areas of primary care. These four would benefit from Lia, but they are still ideas.
Childhood vaccination
Routine childhood vaccines
Lia would remind parents about a late vaccine and point them to the UBS to get it.
58,32%was the coverage of the second dose of the MMR vaccine in Tocantins in 2025, against a target of 95%Partial data up to July. Brazilian Ministry of Health, Nota Técnica Conjunta nº 368/2025
Syphilis in pregnancy
Testing and treatment during prenatal care
Lia would remind the pregnant woman about testing and treatment, in confidence. The partner would be contacted by the team, never by Lia.
17.8 per 1,000live births with congenital syphilis in Tocantins in 2024, the highest rate in the countryBrazilian Ministry of Health, Syphilis Epidemiological Bulletin, 2025
HPV vaccine and cervical cancer
Cervical cancer prevention
Lia would invite people for the HPV vaccine and the Pap test, without ever giving test results.
83,91%was the HPV vaccine coverage among boys in Tocantins in 2025, versus 95.21% among girlsSES-TO, via CONASS, 2026
Child health follow-up
Baby and child checkups
Lia would remind parents of the baby's checkups at the UBS. Fever and alarm signs would go straight to the team.
11.16 per 1,000live births was the infant mortality rate of Araguaína in 2023IBGE Cidades
Partners
LUMEN Oncologia e Mastologia
Araguaína (TO)
“The light that inspires trust. The care that transforms journeys.”
A partner of Mama em Dia, the planned study to support women through breast cancer screening.