BLOG: From Hype to Impact – What Will It Take for AI to Truly Transform Global Health?
By Anna Frellsen, CEO, and Mohit Mathur, CTO at Maternity Foundation.
AI is everywhere in global health today — headlining conferences, dominating strategy decks, and driving funding pitches. And rightly so. The potential is enormous: AI can help break down barriers in hard-to-reach areas, personalise diagnostics, transform care delivery, and revolutionise training of healthcare professionals.
Yet in low-resource settings — where the need is greatest — scalable, effective AI solutions remain elusive. While AI technology is advancing rapidly, its potential to create ethical, equitable, and meaningful impact in these contexts continues to lag behind.
What’s Needed for AI to Deliver Real Impact
To ensure AI meaningfully improves health outcomes in low-resource settings, we believe several key actions are needed:
- Design for offline use: Connectivity is not a given. AI solutions must work in environments with limited or no internet access.
- Develop human-centered tools that support — rather than replace — healthcare professionals.
- Make it accessible: Interfaces must be intuitive, culturally appropriate, and adapted to local languages and needs.
- Develop policy frameworks for the safe and effective integration of AI into health systems.
- Ensure ethical data management: Prioritise ethical data collection, use, and sharing.
These insights are grounded in Maternity Foundation’s experience scaling the Safe Delivery App across Sub-Saharan Africa, the Middle East and Northern Africa, Latin America and the Caribbean, as well as Asia and the Pacific. The app supports midwives and other healthcare professionals in providing quality maternal and newborn care in low-resource settings.
The key to its success? Local ownership, contextual adaptation and system integration. The app is now available in over 35 languages, tailored to national needs and different contexts, and is embedded in education and training programmes through partnerships with governments, UN agencies, and other organisations.
We are currently exploring and testing the integration of various AI-powered services into the app. One of these is the NeMa Smartbot, developed in collaboration with Neuvo Inc. Global, designed to offer faster and more accessible guidance to users when they need it mos
Like the app, the NeMa smartbot is designed to also work offline, ensuring healthcare professionals can use it regardless of connectivity. All responses are scientifically verified and based on audited clinical content provided by Maternity Foundation, while anonymous usage data is collected to help us continuously improve the tool
Testing so far has shown that building smart, memory-efficient AI requires research, rigorous testing, and continuous iteration. We will be embedding these learnings into the next version of our Safe Delivery App.
Three Myths Holding Us Back
We’re encouraged to see more organisations joining us on the AI journey—building, using, and embedding AI tools to support people in need and enhance the quality of healthcare.
To bring even more partners on board, we’re addressing common myths below while sharing key recommendations for the way forward.
Myth 1: “AI Is Something Entirely New”
AI feels novel because it’s changing how we interact with machines — but it’s part of a long tradition of toolmaking. Treating it as something alien fuels hype without understanding, and discourages critical engagement.
Recommendation: Integrate AI into existing platforms. Build on what works. Treat AI as an evolution — not a revolution.
Myth 2: “AI Is Cheap”
AI solutions may appear inexpensive at first, but the real costs lie beneath the surface:
- Data preparation – cleaning, labeling, and validating datasets
- Infrastructure – hardware, cloud services, and ongoing maintenance
- Integration – adapting AI to workflows and systems while ensuring long-term support
- Legal compliance – addressing bias, privacy, and ethical risks
- Third-party platforms – hidden costs, limited customization, and IP concerns
We’ve seen firsthand how costly it is to build and sustain ethical, inclusive, and reliable AI systems. The investment is real — but worthwhile when done right.
Recommendation: Shift the narrative. AI isn’t free — it’s a strategic investment.
Myth 3: “Data Is External and Ethically Neutral”
Data isn’t static or neutral — it reflects the systems and people who created it. Every data set carries values, assumptions, and omissions.
Bias is inevitable. That’s why ethical data governance is essential — from how we collect and prompt, to how we validate and co-create.
Recommendation: Treat data as embedded, not external. Govern it with care, context, and community.
Where Are We on the Hype Curve?
In global health, we are at a pivotal moment. The excitement around AI is high — but now the hard work begins. If we want AI to deliver where it matters most, we must build it differently — and do so with humility, collaboration, and deep local engagement.
There will be trials and errors, successes and failures in developing AI solutions for low-resource settings, but if we stay grounded in the local context and needs, AI can truly transform global health.
