There is a version of this article that leads with a statistic about maternal mortality or financial exclusion in Nigeria. We are not writing that version. Those statistics exist, and they matter, but they are not what this article is about.
This article is about what happens when you decide that the statistics are not enough — and you go to find out what the problem actually looks like in the specific communities, facilities, and systems where you are planning to intervene.
In 2025, CcHUB conducted structured problem-validation workshops across four Nigerian states (Kaduna, Kano, Gombe, and Lagos) as part of a Gates Foundation-funded AI/DPI program implemented in partnership with DIAL and DSN. The program supported nine startups applying artificial intelligence and digital public infrastructure to critical service delivery challenges. Six of them focused on validated health challenges, specifically healthcare financing and maternal/neonatal mortality.
The workshops were the methodological foundation on which the entire program was built.
The Sequence Mattered
The validation process was a two-phase process run in sequence to ensure structured responses.
The first phase was a desk review. Before any engagement with communities or stakeholders in the four states, the team conducted a comprehensive review of existing data, research, and government reports on the challenge areas under consideration — maternal mortality, neonatal mortality, child malnutrition, women’s financial inclusion, healthcare affordability, and agriculture. This produced initial problem framings for each state, grounded in evidence but untested against on-the-ground reality.
The second phase was the quadruple helix workshop. Structured workshops in each state brought together community members, health workers, state government officials, facility administrators, and sector practitioners. The purpose of these workshops was not to confirm what the desk review had found. It was to interrogate it: to validate the problem framings against the accounts of people inside the systems those framings described, and to surface what the data alone could not tell us.
The state selection was intentional. Gombe, Kano, Kaduna, and Lagos were chosen based on a structured evaluation of criteria including sectoral challenge severity, digital readiness, political economy, ecosystem strength, and existing investments in digital infrastructure.
What We Found Across Four States
Each state presented a distinct combination of infrastructure maturity, economic barriers, cultural context, and trust in digital systems. The program developed a structured recommendation for each state and problem area across three dimensions: digital infrastructure availability, economic barriers to adoption, and community trust in technology.
What this workshop revealed was that no two problem areas presented the same conditions, and the innovation approach appropriate for one context was often entirely wrong for another.
Gombe — Maternal Mortality
Gombe presented the most constrained conditions in the portfolio. The state’s maternal mortality ratio — 993 deaths per 100,000 live births in 2020, against a national average of 512 — reflects a health system under significant structural pressure. Only 30% of pregnant women in Gombe had access to professional healthcare services; 70% depended on traditional birth attendants. Of the 14 priority maternal and neonatal health indicators that could be tracked through facility-based data, 12 were nominally in the state’s DHIS2 system — but data completeness was below 60% at least 40% of the time, with under-reporting ranging from 10% to 60% depending on the indicator.
The infrastructure classification for Gombe maternal mortality was the most challenging in the program: low digital infrastructure, high economic barriers, and low community trust in technology. The practical implication for any startup proposing an AI or DPI-based solution was significant. A product that depended on consistent data feeds from the health system, or that assumed community willingness to engage with a digital platform, was not a viable solution for this context in its current form. The validated approach for conditions like Gombe’s is offline-first, community-mediated, and low-tech in its user interface even when the underlying system uses sophisticated data processing.
Kano — Child Malnutrition and Women’s Financial Inclusion
Kano presented two distinct problem areas, each with a different root cause structure and a different infrastructure picture.
On child malnutrition, 53.1% of children under five in Kano showed stunting — among the highest rates in the country. The drivers were interconnected: soil degradation, erratic rainfall, limited access to quality agricultural inputs, poor post-harvest storage infrastructure, and cultural feeding practices that limited dietary diversity. The data picture was clearer than in Gombe — Kano’s digital and agricultural infrastructure sits at a medium level — but the solution pathway required integrating climate and geospatial data (including NiMet meteorological data) with community-level agricultural support, not simply building an information platform.
On women’s financial inclusion, the findings were more culturally specific. Many women in Kano operate through informal savings groups (locally known as Ajo) which function as trusted financial mechanisms precisely because they are not formal or digital. The infrastructure classification was medium digital infrastructure, medium economic barriers, medium trust — but the “medium trust” finding masked significant variation. Trust in formal digital financial systems was considerably lower than the aggregate figure suggested, particularly among women who had not previously engaged with formal banking. Hybrid models — offline-plus-online, with strong community intermediary roles — were identified as the appropriate innovation approach.
Kaduna — Neonatal Mortality and Women’s Financial Inclusion
Kaduna’s neonatal mortality challenge is shaped heavily by cultural and behavioural factors that sit upstream of the healthcare system itself. Cultural beliefs that delay early neonatal interventions — including practices around the timing of first contact with formal health facilities after birth — contribute to preventable deaths that digital health records alone cannot address. The infrastructure classification was medium digital infrastructure, high economic barriers, and medium trust in technology.
The implication for startups was that solutions focused purely on data and clinical workflow would miss a significant portion of the problem. Community engagement and behaviour change components were not optional additions to a health technology product in this context — they were core design requirements.
On financial inclusion, women in Kaduna face a compounding constraint: low digital literacy, limited mobile phone ownership among women specifically, and dependence on informal savings mechanisms that exist because formal alternatives have not been accessible or trustworthy. The validated innovation approach for this context was USSD-based and agent-assisted — technology that works on basic phones without requiring digital literacy at the user level.
Lagos — Healthcare Affordability and Financing
Lagos presented the most favourable infrastructure conditions in the portfolio: strong digital infrastructure, high trust in technology, and high economic barriers. The healthcare affordability challenge in Lagos is not primarily a data problem or a technology access problem. It is a financing problem. Low-income populations face high out-of-pocket healthcare costs with limited insurance penetration, and the distrust that exists is not in digital systems generally but specifically in digital financial products applied to health. This has seen significant mis-selling and failed product launches in the Nigerian market.
The validated approach for Lagos was more sophisticated: AI-driven healthcare financing tools were viable here in a way they were not in Gombe, but they required explicit trust-building mechanisms and a clear demonstration of how they differed from previous products that had not delivered on their promises.
What the Classification System Produced
One of the most operationally useful outputs of the validation process was a structured framework for matching innovation approaches to context conditions. The program developed a priority score for each state and problem area based on five dimensions — digital infrastructure, economic barriers, prototyping locations, trust in technology, and cultural sensitivity — and used that score to define what type of innovation approach was appropriate for each context.
This framework prevented a category of mistake that is common in technology-for-development programs: applying the same product architecture to contexts that require fundamentally different approaches. A solution designed for Lagos’s high-infrastructure, high-trust environment cannot be deployed in Gombe without redesign. Building that insight into the program’s architecture from the start — before startups were selected and before any product development began — changed what the open call asked for and what solutions were considered viable.
The Decision Framework That Resulted
The validation research did not just produce qualitative insights. It produced a structured assessment framework — scoring each state and challenge area across digital infrastructure readiness, economic barriers, prototyping viability, technology trust, and cultural and religious sensitivity — that directly determined what the programme’s open call asked for.
Startups were not selected on the strength of their technology. They were selected on the degree to which their startup could respond to validated problems and was designed to operate within the infrastructure, affordability, connectivity, trust, and institutional realities of the communities. In contexts classified as low-infrastructure and low-trust — Gombe, primarily — the assessment framework explicitly identified offline-first, USSD-based, and agent-assisted models as the appropriate innovation approach. In Lagos, AI-driven financing tools were viable but required trust-building mechanisms built into the product design, not added as a communication strategy after launch.
This is what Human-Centred Design does when it is operationalised as a program methodology rather than invoked as a principle. It changes the selection criteria, changes the product brief, and changes what counts as a viable solution before any investment in building begins. The cost of that rigour is time and the willingness to let research findings disrupt comfortable assumptions. The benefit is a significantly reduced probability of funding solutions that solve the wrong version of the problem.
What the Startups Built and What Happened Next
Nine selected startups focused on health, financial inclusion, and agriculture across the program’s target states. Within three months of the pilot phase, their impact was noteworthy.
MyItura, working on healthcare financing in Lagos, onboarded 45 healthcare providers specifically for its Mediloan product, processed over ₦47 million in healthcare financing requests, and disbursed over ₦10 million to more than 3,250 users. Its near-term targets include scaling to 1,000 healthcare facility integrations, expanding access channels across mobile, web, USSD, WhatsApp, and IVR, and growing cumulative financing requests from ₦50 million to over ₦4 billion.
Eight Medical, operating across three LGAs in Gombe State and expanding into Lagos, Edo, Kebbi, and Adamawa, onboarded over 1,500 users post-pilot, registered 18 healthcare facilities, and completed over 1,000 emergency transport rides with 40 transport workers onboarded. The startup is completing Lagos State integration in partnership with LASHMA, the Ministry of Health, and the Lagos State Health Service Commission, and has World Bank-confirmed pilots active in Edo, Kebbi, and Adamawa. Its target is ten states with active deployments, with at least two featuring government health insurance fully integrated as a transport subsidy layer.
Flolog, focused on maternal health, onboarded over 600 pregnant women, supported three live-birth interventions during its MomCare pilot, and activated over 20 local health agents.
XchangeBox (operating as Kidashi), working on asset financing for rural women traders, disbursed ₦2 million worth of asset financing in the first month of piloting alone, provided small business financing to over 94% of users onboarded, and recorded a 76% repayment rate. Its expansion targets include onboarding 500 rural women traders across three LGAs, securing seed funding of $300,000 to $500,000, and launching agent networks in two additional states.
Evet, supporting smallholder livestock farmers, built a platform serving 1,064 farmers and 99 extension agents during the programme, with near-term targets of 100,000 verifiable and traceable livestock farmers, 5,000 extension agents, and over $10 million in market offtakes for livestock across five additional northern states.
UHC Easy Cover onboarded over 150 persons during the pilot phase, targeting 100,000 new enrollees over the next two years.
Across the portfolio: more than 5,900 cumulative users, 63 healthcare facilities reached, ₦47 million in healthcare financing requests processed, ₦10 million successfully disbursed, 1,000 emergency rides completed, and three live-birth interventions during the MomCare pilot. Two startups have received Social Impact Fund recognition, one is in active partnership conversations with the World Bank, and several are in advanced discussions with development finance institutions.
What This Means for Those Designing Similar Programs
The HCD-driven validation process that underpinned this program took weeks, required physical presence in four states, and produced findings that were uncomfortable in some cases. This is because they showed that the most needed solutions were not always the most technically interesting ones, and that the communities most affected by the problems the program targeted were also the communities least served by the technological approaches that were easiest to build.
Three things stand out as most transferable to others working in this space.
The desk review and the stakeholder engagement are not interchangeable. The desk review produces an assumed problem framing. The stakeholder engagement validates it. Running one without the other produces either unvalidated assumptions or conversations without an analytical frame to structure them. Both phases are necessary, and the sequence matters.
Infrastructure classification is a design input, not necessarily a background condition. Knowing that a target community has low digital infrastructure and low trust in technology is only useful if it directly shapes what you ask innovators to build. If that classification sits in a research report without influencing the open call problem statement and the startup selection rubric, the program results might not be meaningful. Ultimately, the validation work would not have been operationalised.
The most important findings are often the cultural and behavioural ones. The data gaps, the infrastructure deficiencies, and the economic barriers were visible in existing research. What the workshops surfaced that the desk review could not was the trust landscape, the role of informal institutions like Ajo savings groups, the cultural practices that shape health-seeking behaviour, and the specific ways in which previous technology interventions had failed to earn community adoption.
If you are designing an AI or DPI program in an African health or financial inclusion context and want to partner with us, kindly reach out to us.