Offline-First AI: A Mandate for Resilient Healthcare in Nigeria

The article argues that cloud-dependent Artificial Intelligence (AI) solutions are fundamentally ill-suited for the realities of Nigerian healthcare. Citing personal experiences in Lagos hospitals, the author highlights the critical unreliability of internet connectivity and power supply across much of the country. This instability renders cloud-based diagnostic tools or clinical decision-support systems ineffective precisely when they are most needed, posing a direct risk to patient safety and care continuity.

Beyond connectivity, significant concerns arise regarding data privacy and cost. Nigeria's National Health Act and Data Protection Act 2023 mandate strict confidentiality and control over patient information, with restrictions on data movement outside the country. Cloud AI, often hosted abroad, complicates compliance and exposes sensitive health data (e.g., HIV status, mental health) to foreign jurisdictions. Furthermore, the variable, dollar-denominated costs of cloud services are unsustainable for local primary healthcare centers operating on fixed, Naira-based budgets.

The solution, the article posits, lies in an "offline-first" approach. This involves deploying open-source AI models like Llama, Mistral, and Gemma on local servers within hospital premises, ensuring data remains on-site and systems function independently of external internet. This model has already proven successful in resource-constrained settings globally, such as Hikma Health's EHR system for displaced populations in Lebanon and Nicaragua, and in specific Nigerian hospitals that have adopted local EMR solutions.

Implementing offline AI is not without its challenges. It requires upfront investment in hardware, local technical expertise for maintenance, and effort to train or adapt models on local clinical data. Additionally, compliance with Nigerian data protection laws still applies, requiring local registration and data protection officers. However, these challenges represent necessary investments in building resilient, locally controlled health technology.

The core message is a call for African builders to adopt offline-first as a foundational design principle, rather than a post-failure workaround. By owning the AI models, servers, and data, Nigerian healthcare can leverage AI effectively, tailored to its unique infrastructure and regulatory environment, without waiting for external permissions or perfect connectivity.