Introduction

Health Systems Strengthening Cohort has delivered breakthrough advances in child well‑being, health, and protection through responsible AI or AI-driven systems deployed in schools, helplines, clinics, hospitals, and community settings.

 

Their work demonstrates a clear pattern: open, ethically governed AI—co‑created with frontline healthcare and child protection practitioners—can deliver safe, scalable, and transformative services for children and vulnerable populations.

The combined portfolio shows measurable progress in prenatal clinical care, children’s comprehensive well-being, child protection and has strengthened AI capabilities in their respective local innovation ecosystems. 

Impact By the Numbers

The cohort’s AI-enabled solutions are now supporting tens of thousands of children across schools, clinics, and helplines. The deployments include: 

• 10,000+ children in 30+ schools in Bangladesh
• 1,500+ pregnancies supported with clinical decision assistance in India
• A national child helpline center piloting real-time AI transcription & triage in Tanzania
• First pediatric fever chatbot in Spanish, enabling safe guidance across Colombia

They demonstrate that AI can operate safely in high stakes and real world environments. 

Key Insights from the Cohort

1. Why Startups/Innovations Matter in Overstretched Public Systems

According to Anthropic, actual large language models’ job coverage of healthcare practitioners, social services and healthcare support remains much lower at 4-8% than the theoretical share of job tasks that LLMs could perform at 38-58%.

Yet, global health and child protection systems continue to face significant resourcing constraints, with Child Helpline International reporting that 14% of all non‑counselling contacts in 2024 going unanswered due to insufficient staffing and the World Health Organization estimates a projected global shortfall of 11 million health workers by 2030, with shortages affecting countries at all income levels due to chronic under‑investment in training, retention, and health‑system capacity. This underscores a widening treatment gap that traditional in-person consultations and voice‑based helpline models cannot close alone.

While in person consultations and visits will always remain a critical part of healthcare and child protection, a rapid digital pivot is reshaping how young people seek support: voice calls are still the biggest share, and there is a sharp rise in nonverbal channels such as chat, SMS, and messaging apps. 

Multichannel digital services—like the chat based offerings from private players such as Moner Bondhu, BITZ, and Docokids—can allow national and public healthcare services and helplines to manage caseloads more efficiently, reduce waiting times, and engage youth and young parents with sick children who prefer or feel safer using nonverbal and faster communication. 

Together, these trends highlight that expanding digital access is one strategy for narrowing the treatment gap in the face of limited human resources.

2. High‑stakes decisions demand human oversight and responsible AI

The cohort operates in the healthcare and child protection space where mistakes can be irreversible and cause physical/emotional harm or lead to delays in seeking help and even life‑threatening outcomes.

If the system mishandles or improperly stores health disclosures or personally identifiable information, especially for minors, it may violate privacy expectations or national legal frameworks. In regulated fields such as health, factual errors can lead to compliance violation, resulting in penalties, customer harm or litigation. 

Operating in high-stake and sensitive domains across the cohort—from fever guidance to fetal monitoring to child protection, the teams prioritize human expert-led (be it social services, medical and clinical) oversight, explainability, patient and medical ethics alignment, data protection and more importantly, human-in-the-loop workflows.

The experiences of the cohort confirm due to the higher stakes and sensitive domains, responsible AI practice and human in the loop are non‑negotiable pillars when working with children and vulnerable groups.

👉 When the stakes are high, involve human experts in the loop and uphold high responsible AI practices. 

3. AI should support clinicians and frontline workers, not replace their decision-making

AI outputs can appear authoritative while being clinically invalid, creating safety risks if used without human oversight.

Docokids’ solution is designed to route high-risk medical cases to the care they need, while easing the pressure caused by low risk cases that would otherwise occupy national healthcare systems and emergency wards intended for truly critical patients. In the words of Docokids’ CEO Natalia Cano, her vision of their AI-powered tool is to:

4. Strongest predictor of scale: AI built by and with practitioners

Frontline workers—GPs, counselors, pediatricians, child helpline managers—are now co-designing and/or delivering AI-driven solutions. These practitioner-led systems are critical to ensure that the solutions fit real workflows, reduce administrative burden, improve triage and increase adoption and long-term viability. 

Half of the cohort, Moner Bondhu and Docokids, are medical practitioners themselves ,i.e. Pediatricians and counsellors in their respective countries, Bangladesh and Colombia. The other half of the cohort, CareNX and BITZ, work endlessly with their partner practitioners in the field i.e. GPs in India and child helpline operators in Tanzania. Building by if not alongside practitioners also helps to mitigate symptomatic bias or solution bias. 

👉 This is, the opinion of the Ventures Fund, one of the strongest predictors of sustainability and scale. When possible, future funding should prioritize practitioner-led or involved projects.

5. Data is the bottleneck—Not AI models

Across the cohort, when baseline data and established clinical guidance did not exist, data was the main project bottle. Ethical and comprehensive data collection is also underestimated in cost, time and resources. 

In addition, they were also working with resource-stretched practitioners such as General Physicians and child helpline call operators, that had to set aside time to understand and be trained in the data collection procedures and facilitate the data collection. 

Sensitive data requires additional and extensive approvals (i.e. MOUs, ethics, privacy protocols) and procedures (i.e. Anonymization, generation of synthetic data based on actual data). If the data exists, the quality of data may also vary.

Data scarcity, layered with systemic issues, also appears to continue to be a recurring issue in that we see.

👉 Invest both time and resources to get annotated quality data.

👉 Future funding must prioritize data collection (if baseline data is scarce or established guidelines are non-existing), data governance, shared datasets, and ethical collection pipelines.

Looking Ahead

Building on the cohort’s results, UNICEF Venture Fund will continue to back frontier technology innovations that demonstrate responsible, practitioner‑led deployment in real‑world health and child protection systems. We will prioritize solutions that strengthen overstretched public systems by augmenting—not replacing—frontline workers; embed human oversight and ethical governance from day one; and invest early in the data, partnerships, and safeguards required for safe scale. 

As AI adoption accelerates, our focus remains clear: Supporting startups that can translate technical promise into measurable outcomes for children and vulnerable populations, especially in low‑resource, fragile, and emergency settings where the need is greatest.

Follow the journeys of our companies as they continue to translate responsible AI into real‑world impact for children and vulnerable communities:

• BITZ IT (Kenya) - AI and ML-powered management system to support call centers assisting survivors of gender-based violence and violence against children
• CareNX (India) - Fetal heart monitor system that integrates an AI algorithm and off-shelf devices to tackle the issue of scarcity of gynecologists in rural areas
• Docokids (Colombia) - AI-driven Fever Module to enhance support for parents in remote and rural areas and improve the efficiency of the pediatrician's work

• Moner Bondhu (Bangladesh) - Accessible holistic well-being services — in-person and online — for children and young adults