The challenge
Community health workers (CHWs) are the first link in primary health care across Africa. In rural areas, they are often the only health contact a family ever has.
Their effectiveness depends on how regularly they are trained, and that is where health systems struggle: in-person training is expensive, logistically heavy and reaches only a few dozen workers at a time. Regular national coverage is beyond available budgets.
Digital delivery is the obvious answer, but it runs into a constraint that online solutions cannot engineer around: connectivity. Patchy network coverage, the cost of mobile data, digital literacy: a connected solution reaches the best-served workers and misses everyone else.
Source: Kajou field observations in Senegal.
Our solution
kSANTÉ is a training and practice-support app for community health workers. It runs entirely without an internet connection and combines three components, delivered on a microSD card inserted into the health worker’s smartphone.
On video
Kevin Abou Khalil, Kajou’s Director of Development, presents the solution and how it works in the field.
Watch the videoTrain and equip dispersed, poorly connected health workers that online programmes do not really reach.
An AI assistant and content built on the ministry’s protocols, kept up to date and available anywhere, even without a network.
Continuous visibility on real usage and on health workers’ skills, and a direct channel to communicate with them and send them training content and information in real time.
Rich content and AI with no connectivity costs, including in the most remote areas.
Open architecture, content based on national protocols, progressive transfer to the State under a Build-Operate-Transfer model.
Designed to bring together the existing e-learning tools of ministries and NGOs, and to work alongside other e-health tools.
Low-cost distribution on microSD cards and centralised management without heavy IT infrastructure.
A major innovation
Conventional generative AI requires constant connectivity and heavy computing power, both scarce in rural Africa. kSANTÉ reverses the logic: answers are produced and validated upstream, then embedded on the card.
No text is generated on the phone. A lightweight semantic engine instantly retrieves, offline, the validated answer closest to the health worker’s question.
From national guidelines and protocols, around 15,000 question-answer pairs are generated, then reviewed and validated by medical experts.
The database is loaded onto the Kajou microSD card, inserted into the health worker’s smartphone.
An on-device classifier maps the question, asked in plain language, to the most relevant validated answer. No internet needed.
of offline AI answers rated relevant by health workers
[Exact source and scope to be confirmed]
No invented answers. Every answer is written in advance from national protocols and validated by experts.
Practice support, not a diagnostic tool. In case of doubt or danger signs, the health worker refers the patient to the health post.
Accessible. Answers in plain French with an audio version; Wolof and other local languages are being integrated.
No patient data. The assistant collects no information about patients.
The AI assistant is currently deployed in Senegal and will be activated in new countries as deployments progress.
An open infrastructure
Once health workers are equipped, kSANTÉ becomes a channel to deliver new training quickly, with no new hardware and no new distribution campaign.
With the support of Being and Grand Challenges Canada, we are deploying a mental health component to community health workers already equipped with kSANTÉ in Senegal: new training modules, and an AI assistant trained to answer their questions on these topics. Other themes are in progress, such as parenting and non-communicable diseases.
Our digital learning studio in Dakar helps you turn it into interactive modules, in French and local languages, deliver it on the cards already in circulation and train the AI assistant on your guidelines. The dashboard then measures usage.
Propose a training programmeTheory of change
INPUTS
OUTPUTS
OUTCOMES
For health workers
Stronger knowledge, faster and more relevant care, greater motivation.
For governments and NGOs
Visibility on usage and skills, better-informed decisions, a direct channel to health workers.
For families
Better knowledge of their health and faster referral to the health post.
IMPACT
Improvement of key public health indicators, particularly maternal and child mortality.
Questions asked by health workers and usage data flow back to the teams and continuously improve content and the AI.
Deployment
kSANTÉ is deployed with Ministries of Health and nine partner NGOs. Select a country to see where deployment stands.
kSANTÉ is built on an open architecture. The ministry does not rent a proprietary service: it progressively acquires a sovereign digital asset that it ends up owning.
Kajou builds and operates the system alongside the ministry in the first years, then transfers the core components, governance, content and dashboard, while training national teams.
kSANTÉ is designed to work with national health information systems and other e-health tools: as a standalone app, inside a super-app, or by taking over data collection functions already in place.
At each stage, the number of health workers equipped grows, the evidence of impact gets stronger and the ministry takes over the system.
PHASE 1 · 6 TO 12 MONTHS
We measure the effect on health workers’ knowledge and practices, adoption and cost, and build the relationship with the ministry.
PHASE 2 · 18 TO 24 MONTHS
A national version of kSANTÉ is created, the transfer begins and impact is evaluated on public health indicators, across entire regions.
PHASE 3 · ABOUT 2 YEARS
All active health workers in the country are equipped. kSANTÉ becomes a national platform under public ownership.
Our goal: 100,000 community health workers equipped in Africa by 2030, serving 30 million people.
Deploy kSANTÉ in your countryEvidence from evaluation
From January to June 2026, an independent randomised controlled trial, conducted by Eval-Lab and funded by the Fonds d’innovation pour le développement, followed 570 community health workers in the Saint-Louis region of Senegal.
After a baseline questionnaire, a public lottery assigned health workers to three groups of about 200. Three months later, all of them took the same knowledge test.
In standard deviations, against reference benchmarks for education interventions (Kraft, 2020)
RESULT 1
After the initial novelty peak, usage settles at a durable level instead of collapsing. That is not a given: rapid abandonment is the most common failure mode of such programmes.
RESULT 2
+0.53 points out of 20 for equipped health workers, +0.75 with calls. The gain is concentrated on the topics covered by the modules, and is larger among workers who started furthest behind. Calls add no detectable effect: supervision by the head nurse of the health post structures usage better.
RESULT 3
Equipped health workers report holding more community talks and referring more people to the health post. An unanswered question no longer means a trip and days of waiting.
Before kSANTÉ I had to travel to come and ask the ICP the question, which was not easy. But now I get the answer straight away.
I use it to prepare fieldwork: I watch the video first, and then the AI assistant for extra questions.
I ask the AI the question and show the answer to the pregnant woman to back up what I am saying.
Impact report
Trial design, detailed results, analysis by department and qualitative interviews with health workers: read the full evaluation conducted by Eval-Lab in Senegal.
Read the full reportThey trust us



























kSANTÉ received the award for best AI social innovation from Meta’s Llama Impact Accelerator.
Contact
Ministry, NGO, funder or health operator: would you like to deploy kSANTÉ in your country, adapt it to new languages or deliver your own training content through it? Let’s talk.
The kSANTÉ team will get back to you shortly.