![Boyer, color aquatint [background] with additional hand-coloring [figure] by Louis François Charon](../images_blogs/art/ai-opportunities-haiti-2026.jpg)
Any plan for AI in Haiti has to start from the conditions on the ground. Since President Moïse was assassinated in 2021, armed groups have taken control of stretches of Route Nationale 1 and 2 and parts of Port-au-Prince. The 2010 earthquake killed an estimated 160,000 to 230,000 people and caused about $8 billion in damage. The 2021 earthquake near Les Cayes killed about 2,200 and caused about $1.5 billion more. Hurricanes and floods arrive most years.
Those facts decide which tools can work. A tool that needs fixed broadband, steady power or an office IT department will fail. A tool that runs on a phone, in Haitian Creole, and still works on a weak signal has a chance.
The assets are real too. About 80% of Haitians have a mobile phone, and Digicel Haiti and Natcom cover most of the country. Remittances bring in about $3.8 billion a year, among the highest shares of GDP in the Americas. Roughly half the workforce farms, the median age is about 24, and a large, skilled diaspora in Miami, New York, Boston, Montreal and France wants to help.
Remittances: cut the cost of sending money home
Remittance fees are the clearest place where software saves Haitian families money. The main channels, including Western Union, MoneyGram, CAM Transfer and Unitransfer, have charged 7–10% of the amount sent, plus an exchange-rate margin. On a $300 monthly transfer that is $21–30 in fees before the exchange margin.
Part of that cost is compliance and fraud risk in a corridor that banks treat as high risk. Transaction monitoring models help here. A family support transfer has a recognisable pattern: the same sender, the same recipient, similar amounts at regular intervals. A model that knows that pattern flags the unusual transfers for review and lets the ordinary ones through, which lowers the cost of serving the corridor. Digital operators such as Remitly and Ria already compete on this basis.
The arithmetic is simple. Each percentage point of fees on $3.8 billion is $38 million a year. If half of all transfers moved from a 9% total cost to 4%, Haitian households would keep about $95 million a year.
Fonkoze, Haiti's largest microfinance institution, has branches in all ten departments and years of experience in digital financial services. It is the obvious local partner for a lower-cost receiving service, with fraud models tuned to Haitian family transfer patterns.
Agriculture: prices, plant disease and export records
Smallholders grow rice in the Artibonite Valley, food crops on the Northern Plain, and coffee, cacao and mangoes on the southern peninsula. Most of them sell without knowing the price in the next market, get little extension support, and farm eroded hillsides. Three phone-based tools address those gaps.
- Market prices by WhatsApp and SMS. Haiti's national food security coordination body, CNSA, monitors prices in the main markets. Sending those prices in Creole to farmers in Croix-des-Bouquets, Saint-Marc, Cap-Haïtien and Gonaïves gives them a number to bargain with. It costs the farmer nothing to use.
- Crop disease diagnosis from a photo. PlantVillage, built at Penn State University and widely used in sub-Saharan Africa, diagnoses cassava, maize and other crop diseases from a phone picture. Haiti needs versions trained on yam, sweet potato, malanga, banana and plantain, with a Creole interface.
- Traceability for coffee and cacao. Specialty buyers want farm-level records of origin, grower and processing. Cooperatives such as COOPCAB in coffee could capture GPS and processing data by phone. Traceable specialty coffee can sell for roughly $5–8 a pound against $1.50–2.50 for undifferentiated beans.
Haiti already exports products with a premium story: Haitian Bleu coffee, Grand'Anse cacao and Francisque mangoes. What the cooperatives often lack is the paperwork that proves it.
Disasters: faster damage maps and safer aid routes
Haiti moves from one recovery to the next, so tools that shorten each recovery pay off repeatedly.
Damage assessment from satellite images. After 2010 and 2021, UNOSAT and the Copernicus Emergency Management Service mapped building damage from satellite imagery, largely by hand over weeks. Models trained on those same Haitian images can now produce a first damage map in 24 to 48 hours. The World Bank's Global Facility for Disaster Reduction and Recovery (GFDRR) has funded this kind of work, and Haiti's civil protection agency, the DPC, is the natural user.
Flood and landslide warnings. Heavy rain on Haiti's bare hillsides turns quickly into flash floods and landslides. Rainfall forecasts combined with elevation data and past landslide records could give communities in Artibonite, Grand'Anse and Nippes 12 to 24 hours' notice.
Aid routing. ACAPS and OCHA track road access and areas controlled by armed groups. Combining those maps with population data lets agencies route deliveries around blocked roads and see which communities nobody has reached.
Education: Creole first, on the phone
Haitian Creole is the first language of almost every Haitian. French is a second language for an educated minority, and English matters mainly in diaspora families. Most education technology used in Haiti still runs in French, which shuts out many of the children who need it most.
Violence and displacement have closed schools across the Port-au-Prince area and beyond, and hundreds of thousands of children are out of school. Short, audio-led lessons in Creole for maths, reading and science, sent over WhatsApp at low data cost, can reach them through a parent's phone. Diaspora teachers, of whom there are many among the more than one million people of Haitian descent in the United States, could add live tutoring sessions once the basic lessons exist.
Health: decision support for community health workers
Haiti has one of the lowest ratios of doctors to population in the Americas, and most doctors work in Port-au-Prince and a few secondary cities. In rural areas, community health workers (agents de santé communautaire) provide most care with little specialist backup.
A structured decision tool on a phone can help a health worker tell a child with uncomplicated malaria, which can be treated in the village, from one with danger signs who needs referral now. Zanmi Lasante, the Haitian partner of Partners In Health, runs hospitals and clinics in the Central Plateau and Lower Artibonite and has the health worker network and community trust to test such a tool properly.
Maternal health is the other priority. Haiti's maternal mortality ratio is the highest in the Americas, several times that of the Dominican Republic or Jamaica, and most deaths come from haemorrhage, infection and pre-eclampsia. A simple risk score at antenatal visits, using age, number of previous births, blood pressure, haemoglobin and obstetric history, tells health workers and matrones (traditional birth attendants) which women should deliver in a facility.
What makes these projects workable
Every application above runs on a phone, works on a weak signal, needs no reliable power at the user's end, and has a Haitian institution able to carry it: Fonkoze for money, CNSA and the cooperatives for farming, the DPC for disasters, Zanmi Lasante for health. The only figure here I would stake money on is the remittance saving, because it is simple arithmetic on known flows. The rest depends on adoption, and adoption in Haiti depends on security more than on software.
The weakest link in this plan
Every recommendation assumes that staff, farmers and health workers can move around enough to use and maintain these tools. In the areas armed groups control, that is not true today. A crop app does not help a farmer who cannot reach the market, and a triage tool does not help if the referral hospital is behind a roadblock. I do not have a good answer for those areas, and projects should say plainly which communes they can and cannot serve.
What to do next
- Fonkoze and a diaspora fintech partner: agree a pilot of a lower-cost receiving service on one corridor, such as Miami to the Nord department, and publish its all-in cost against the World Bank's corridor figures after six months.
- IICA's Haiti office and a mobile operator: run a 60-day WhatsApp price and plant disease pilot with farmer groups in the Artibonite Valley, and measure the prices farmers get before and after.
- Zanmi Lasante: test a community health worker triage tool with a few hundred health workers at existing sites, measuring referral accuracy before any wider roll-out.
- The DPC: sign standing data-sharing arrangements with UNOSAT and Copernicus now, so a first damage map arrives within 48 hours of the next earthquake or hurricane.
- Diaspora engineers: build Creole-first content, starting with 50 short audio maths lessons for primary pupils, and hand them to a Haitian school network to distribute.
Frequently Asked Questions
How can I check the full cost of sending money to Haiti?
Use the World Bank's Remittance Prices Worldwide database, which lists fees and exchange-rate margins by provider for corridors such as the United States to Haiti. Compare the total cost, fee plus exchange margin, for the amount you actually send, because a zero-fee offer can hide a wide margin. For reference, the UN Sustainable Development Goals target an average remittance cost below 3% by 2030.
Who regulates money transfer and mobile money services in Haiti?
The Banque de la République d'Haïti (BRH), the central bank, supervises banks, money transfer companies and mobile money. A fintech that wants to pay out in Haiti needs either its own authorisation or a licensed local partner, which is one reason partnering with an existing institution such as Fonkoze is faster than starting from scratch. Ask the BRH for its current rules on agents and electronic money before designing the product.
Is AI good enough in Haitian Creole yet?
For general text, major translation services and large language models handle Haitian Creole reasonably well, but quality drops on medical, agricultural and legal vocabulary. Every health or farming message should be checked by a native speaker with subject knowledge before it goes out. The Akademi Kreyòl Ayisyen, set up in 2014, is the reference body for standard spelling, which matters when you train or test a model.
Should a tool for rural Haiti use WhatsApp, SMS or voice?
Use the channel your users already have. WhatsApp needs a smartphone and data, which suits market prices and photo-based crop diagnosis in areas with 3G or better. SMS works on any phone and suits short alerts such as flood warnings. For people who read little, recorded voice messages or phone menus reach further, and many projects need two channels, not one.
How can a diaspora professional help without running a project on the ground?
Work through a Haitian institution that already operates there, such as a microfinance lender, a cooperative, a school network or Zanmi Lasante, and offer a specific skill: building a model, writing Creole content, or reviewing data. Avoid setting up parallel field operations, which are hard to secure and duplicate local capacity. Remote tutoring, code and content can all be delivered from abroad.