Jacaranda health has launched an AI-powered SMS system for pregnant women

An AI-powered SMS system has been launched by Jacaranda health to reduce the infant mortality in Kenya which currently stands at 31.771 deaths per 1000 live births.

The system comes only days after the World Health Organisation (WHO) revealed that infants born in sub-Saharan cities surprisingly have an increased mortality rate in their first month of life compared to babies born in rural areas.

Kenya, ranked 15th in the world's highest mortality rate desperately needed a free service to help expectant mothers obtain accurate information, stated Nick Pearson, Jacaranda co-founder.

“Tech-heavy solutions look good, but once they are embedded they might not work so well. While a lot of people have phones with data capability in Kenya, they don’t purchase data so often,” stated Pearson.

Nick Pearson, co-founder of Jacaranda Health Pulse Live Kenya

1 million women have registered to exchange pregnancy-related questions with a team of health advisers via the free service called Prompts.

Prompts uses AI to triage messages and ensure that “danger signs”, including bleeding and loss of baby movement, receive the fastest responses, such as advice to seek emergency care.

Expectant mothers can register for Prompts in English or Swahili at hundreds of health centres across 20 of the 47 counties.

Jacaranda’s help desk receives 3,000 inquiries a day and aims to respond to all messages within 24 hours, while “clinically urgent” queries are addressed within an hour.

Critical messages are flagged to a senior help desk clinician for an urgent response or phone call. For less pressing queries, the AI prepares a canned response that is vetted by the help desk before it is sent.

“The level of confidence in the AI getting the answer right is around 95 per cent, but when it’s potentially life-threatening we don’t want to take that chance,” Pearson says.

“That’s why we keep the human between the patient and the AI. It’s still a human-first system. The AI is serving up a potential canned response. The human agent then checks that it matches the real intent of the question,” Pearson concluded.