Sign Manifesto
Thanks for reaching out!
Oops! Something went wrong while submitting the form.
COMMUNITY ENGAGEMENT
2000+
Health facilities (PHCs) mapped
100
Migrants reached in pilot phase
20+
Languages
Afiya: Reimagining Migrant Health with AI
This case study showcases Afiya, an AI health agent deployed on WhatsApp, created by the UN IOM office in Tunisia and powered by Gooey.AI. It is intended for humanitarian organizations, public health agencies, NGOs, and policymakers working at the intersection of migration and health.
The International Organization for Migration (IOM) is the leading intergovernmental agency on migration, working with 170+ Member States to promote humane, orderly, and cooperative approaches to human mobility. With a presence in hundreds of field locations across more than 100 countries, IOM delivers services ranging from migration health and reintegration support to crisis response.
COMMUNITY ENGAGEMENT
2000+
Health facilities (PHCs) mapped
100
Migrants reached in pilot phase
20
Languages
Theory of Change
The Need for Accessible Guidance on Health Services
Complex health needs and challenges
From January to July 2025, UN-IOM in Tunisia assisted 7,985 migrants and facilitated 1,707 referrals to public and private health facilities. Many migrants arriving in Tunisia present with complex health conditions that are worsened by language barriers, experiences of stigma, and limited familiarity with local health systems. These factors constrain timely and effective access to appropriate healthcare services.
Language and literacy constraints
Migrants frequently lack clear information on where to access diagnostic services such as laboratory tests, X-rays, and medical appointments, leading to delays in care. Physicians experience increased workload pressures, while patients are more likely to miss appointments and face interruptions in treatment continuity.
Language barriers and low literacy levels further restrict migrants’ ability to understand written health information, thus creating barriers to informed health decision-making.
Language barriers and low literacy levels further restrict migrants’ ability to understand written health information, thus creating barriers to informed health decision-making.
Background and Needs
Limitations of paper-based consent mechanisms
Consent from migrants was obtained through paper-based processes that required individuals to read and sign extensive four-page terms and conditions documents. This approach proved inefficient and exclusionary: clinicians frequently declined to participate due to time constraints, while migrants faced the risk of consenting to the sharing of sensitive health data without a full understanding of its implications.
Delays in access to healthcare
Confusion about where to access appropriate services often results in repeated visits to frontline health workers, increasing the burden on both migrants and the healthcare system. Limited access to physicians for follow-up care contributes to elevated health risks among patients.
These delays can exacerbate existing health conditions, reduce the effectiveness of preventive interventions such as vaccinations, and contribute to the progression of chronic illnesses.
These delays can exacerbate existing health conditions, reduce the effectiveness of preventive interventions such as vaccinations, and contribute to the progression of chronic illnesses.
.png)
Interventions & Process
Platform selection and access considerations
In collaboration with the UN IOM Tunisia team, WhatsApp was selected as the primary delivery channel for the AI health agent, Afiya, due to its widespread accessibility among migrant populations, including individuals without local SIM cards. This platform enabled low-barrier access in a context characterized by constrained digital resources.
Technical framework for structured AI development
Gooey supported UN-IOM Tunisia in Afiya's development through a structured, end-to-end AI workflow. The process began with curating a “Golden Q&A” based on the most frequent health-related queries raised by migrants, supported by a verified knowledge base drawing on public health guidelines, and open data sources.
We then configured and iteratively refined prompts to ensure accurate, context-sensitive responses, followed by bulk model evaluations and red-teaming to test performance and potentially harmful queries. The solution was subsequently expanded to multiple languages and deployed first on the web, before integration on WhatsApp.
We then configured and iteratively refined prompts to ensure accurate, context-sensitive responses, followed by bulk model evaluations and red-teaming to test performance and potentially harmful queries. The solution was subsequently expanded to multiple languages and deployed first on the web, before integration on WhatsApp.
Multilingual and low-literacy use
The system was designed to provide 24/7 multilingual support, including regional dialects, and incorporated voice messaging feature to accommodate users with low literacy levels. Additional features included an integrated service locator with over 2,000 GPS-tagged health service locations linked to Google Maps, as well as structured guidance on care pathways. The agent was piloted with 100 IOM beneficiaries, and iterative refinements were made based on user feedback.
Outcomes
Digitizing consent and ensuring ethical data use
In collaboration with the UN IOM Tunisia team, the terms and conditions and consent process were transitioned to a digital format. This enabled migrants to review consent materials in English or French prior to engagement, thereby strengthening informed consent practices. Consent was completed at the outset of interactions, allowing migrants to share personal and health-related information in a safer and more transparent manner.
Integrated health information and service navigation
The intervention incorporated multilingual support, a service locator covering more than 2,000 primary healthcare centers (PHCs), and structured health education content. In addition, Afiya provided clear care pathway guidance to support both migrant and host communities in navigating health services. The system delivered standardized health messaging, directions to vaccination centers, and access to the eVAC platform for vaccination recording.
Adoption and scalability potential
As part of the pilot, the agent was tested with 100 end users, all IOM beneficiaries, who accessed it via printed QR codes available in waiting rooms and at frontline service desks. Based on needs identified through collected interaction data, the agent focused on four key health domains: vaccination, pregnancy, mental health, and access to nearby health services.
Afiya has gained significant uptake and visibility among migrant communities in Tunisia, as well as in neighboring countries and within regional UN IOM offices. The intervention is being actively monitored, with strong potential identified for scaling and replication in other migration contexts.
Afiya has gained significant uptake and visibility among migrant communities in Tunisia, as well as in neighboring countries and within regional UN IOM offices. The intervention is being actively monitored, with strong potential identified for scaling and replication in other migration contexts.
Relevant Links
