Horizon1000: Gates Foundation and OpenAI launch $50M push to strengthen African primary care with AI
The Gates Foundation and OpenAI are funding a new $50 million initiative, Horizon1000, to help several African countries use AI to strengthen primary care. The effort starts in Rwanda and is aimed at protecting essential services after a sharp drop in global health aid last year.
Funding cuts that began in early 2025 led to a nearly 27% fall in development assistance for health compared to 2024. Bill Gates said the goal is to help countries recover ground and ensure new tools reach low-resource settings as fast as they reach high-income markets.
What Horizon1000 will focus on
The program will work with African leaders to identify priority use cases, beginning with Rwanda, which has already set up an AI health hub in Kigali. Paula Ingabire, Rwanda's minister of ICT and innovation, said the focus is responsible use of AI to reduce clinician workload, improve care quality, and reach more patients.
Horizon1000 expects to support up to 1,000 primary health clinics and surrounding communities by 2028. Early targets include care for pregnant women and people living with HIV-especially language support before clinic visits, faster intake on arrival, and cleaner links across patient histories and appointments. As Gates put it, "A typical visit… can be about twice as fast and much better quality."
Why this matters for frontline teams
Some countries have a single doctor for 50,000 people, even in large cities. With fewer health workers and tighter budgets, small gains in speed and accuracy can translate into more people served, shorter queues, and steadier adherence for ANC and HIV programs.
- Less time on paperwork = more face time with patients
- Language support = fewer missed details and clearer consent
- Consistent reminders and follow-up = fewer missed visits and better retention
High-yield AI use cases you can pilot now
- Pre-visit guidance in local languages (SMS, IVR, WhatsApp): basic triage, appointment prep, directions, and consent reminders
- On-site translation: clinician-patient communication across languages, with clear disclaimers and clinician review
- Documentation assist: auto-summarize visits, auto-fill standard forms, and sync to HMIS/EHR where available
- Defaulter tracing and adherence support: reminders, two-way messaging, and simple escalation for high-risk patients
- Protocol checklists for ANC and HIV care: structured prompts tied to national guidelines, always under clinician oversight
- Supply and stock signals: quick flags from routine registers to reduce stockouts
- On-the-job coaching: microlearning for CHWs and nurses embedded in daily workflows
Implementation checklist (start small, measure hard)
- Pick one pathway and one site: e.g., ANC intake at a busy clinic; run for 8-12 weeks
- Define success up front: visit time saved, patient wait time, missed appointment rate, data completeness, provider satisfaction
- Language coverage: prioritize Kinyarwanda, Swahili, Luganda, or local languages; validate outputs with bilingual clinicians
- Privacy and consent: clear patient consent flows, role-based access, on-device redaction for sensitive fields
- Offline-first setup: low-cost Android devices, local caching, sync when connectivity returns; plan for solar charging where needed
- Integration path: FHIR or national HMIS if possible; otherwise CSV exports with unique patient IDs and audit logs
- Governance: risk classification for each use case, human-in-the-loop review, and a reliable fallback (paper or non-AI forms)
Risks and guardrails
- Incorrect outputs: limit scope to approved protocols, require clinician confirmation for clinical suggestions
- Bias and language gaps: test across dialects and patient cohorts; track errors and fix quickly
- Over-reliance: keep decision authority with clinicians; train staff to challenge outputs
- Security: minimize data collection, encrypt at rest and in transit, keep strong device management
Key metrics to track
- Average visit time and percentage time spent on documentation
- ANC 4+ visit completion and on-time first ANC visit
- Viral suppression rate and loss-to-follow-up in HIV programs
- Patient wait time, no-show rate, and referral completion
- Clinician and CHW satisfaction and burnout indicators
What to watch next
Pilots will likely expand from Rwanda to additional countries as partnerships firm up. Expect attention on procurement frameworks, integration with national systems, and standard-setting for safe deployment in low-resource settings.
For national digital health guidance and evidence-based approaches to AI-enabled interventions, see the WHO's digital health resources: WHO Digital Health and Innovation. Rwanda's ICT ministry provides updates on the country's tech infrastructure and initiatives: Ministry of ICT and Innovation.
Building team skills
If your clinic or program is preparing pilots and needs quick upskilling on workflows, prompts, and evaluation, you can explore concise, job-focused AI coursework here: Complete AI Training - Courses by Job.
The bottom line: start with one care pathway, measure relentlessly, keep humans in control, and scale only after the basics-privacy, language quality, and integration-are stable.
Your membership also unlocks:
