I built StarApple AI in Jamaica and have worked on AI projects here for fifteen years. This piece names the five sectors where I think AI will show results soonest in 2026, what to build first in each, and who should own it.
Jamaica's GDP is about $18 billion. Tourism brings in more than $4 billion a year and remittances more than $3.5 billion. Bauxite and alumina, agriculture, energy and financial services make up much of the rest, and the business process outsourcing (BPO) sector has become one of the largest private employers.
Why Jamaica can move faster than most of the region
Jamaica has a National AI Task Force, a tech cluster in New Kingston and Half Way Tree, high smartphone use and a BPO industry that already serves international clients. It also has a small but growing group of local AI practitioners who understand the context. The infrastructure is mostly in place. What holds Jamaica back is that few companies or ministries pick a specific project, fund it and measure it.
BPO: sell AI oversight as a service
The BPO sector employs more than 55,000 people and earns about $700 million a year in exports. AI automation threatens the simplest voice and data-entry work. It also creates demand for people who check AI outputs, handle the cases the software cannot, and supervise automated processes for overseas clients.
Jamaica can market itself as the place where international firms put English-speaking human oversight teams behind their AI systems. That work pays more per hour than routine call handling, and it builds on skills the workforce already has.
Start here: HEART/NSTA Trust and the Business Process Industry Association of Jamaica rework one BPO training track around AI quality assurance, with JAMPRO pitching the result to overseas buyers. Operators such as Conduent, Teleperformance and TTEC, which already test AI tools elsewhere, are natural first partners.
Tourism: forecast demand and sell more on property
Hotels and the Jamaica Tourist Board hold years of booking, spending and complaint data, and little of it is analysed. Demand forecasts built from booking history, flight searches and source-market indicators help hotels set rates. Guest preference models help sell tours, dining and experiences, where margins are higher than on rooms.
Start here: each property turns on the revenue management module already in its property management system and compares rate decisions for one season. The Tourist Board sets up a sector-wide data-sharing arrangement for anonymised visitor data so smaller hotels can use the same forecasts.
Agriculture: satellite crop monitoring and credit for smallholders
Most Jamaican farms are small, grow for the local market or for niche exports such as Blue Mountain coffee, ackee and scotch bonnet peppers, and have little access to formal credit. Satellite monitoring can flag drought and pest stress on specific plots. Scoring models that use production records, weather history and cooperative membership can give lenders a basis to lend to farmers with no credit file.
Start here: the Coffee Industry Board and the Jamaica Agricultural Society run a pilot with Blue Mountain coffee growers covering quality grading by phone camera and traceability records, since that crop already earns a premium that buyers will pay to verify.
Financial inclusion: credit scores for people without a credit file
About 35% of Jamaican adults are underbanked. Many are self-employed traders and gig workers with phones, regular bills and remittance income, but no record at Jamaica's licensed credit bureaus, such as Creditinfo Jamaica and CRIF NM Credit Bureau. Alternative scoring models that use bill payments, remittance history and National Insurance Scheme contributions have been used in Kenya, Ghana and South-East Asia.
Start here: a credit union or microlender runs the model in parallel with its current process for six months, lending on the old rules while recording what the model would have decided, then compares repayment. That shows whether the model works before anyone's loan depends on it.
Public safety: schedule patrols by place and time
The Jamaica Constabulary Force holds years of incident records. Analysis of where and when incidents cluster can guide patrol schedules, violence-interruption work and social programmes without scoring any individual. Person-level prediction, the version that has a record of bias in the United States, should stay off the table.
Start here: the JCF and the Ministry of National Security partner with UWI or UTech to build and test a place-and-time model on the digitised records, with published rules on what data it uses. Twelve months is realistic.
| Sector | First project | Owner | Measure |
|---|---|---|---|
| BPO | AI quality assurance training track | HEART/NSTA, BPIAJ, JAMPRO | Hourly rate of new contracts |
| Tourism | Revenue management module | Hotels, Tourist Board | Revenue per available room |
| Agriculture | Coffee grading and traceability | Coffee Industry Board, JAS | Price per box, rejected lots |
| Financial inclusion | Parallel-run credit model | Credit unions, microlenders | Repayment rate of new borrowers |
| Public safety | Place-and-time patrol model | JCF, Ministry of National Security | Incidents in targeted areas |
Where this plan is weakest
Earlier versions of this piece added up the sectors into a GDP figure. I have removed it: those numbers depended on conversion and multiplier assumptions that nobody has tested in Jamaica. I know which projects to start. I do not yet know how much they will add, and the measure attached to each pilot is how we will find out.
What to do next
- Business owners: pick one workflow you already measure, such as call resolution time, occupancy or loan arrears, and record three months of baseline figures this month.
- Business owners: run a 60-day pilot of one AI tool on that workflow, with one named staff member spending a few hours a week checking its output, then compare against the baseline.
- BPO operators and HEART/NSTA Trust: agree the content of one AI quality assurance course and enrol a first cohort of current agents before the end of the year.
- Credit unions and microlenders: register the data use with the Office of the Information Commissioner, then start a six-month parallel run of an alternative scoring model.
- Jamaica Tourist Board: publish the terms of an anonymised data-sharing arrangement that small hotels can join at no cost.
Frequently Asked Questions
Which law applies when a Jamaican business uses customer data to train or run an AI tool?
The Data Protection Act 2020, which came fully into force in December 2023. It is overseen by the Office of the Information Commissioner, and organisations that process personal data must register with that office. Before feeding customer records into an AI tool, confirm that the purpose matches what customers were told when the data was collected, and check where the tool stores data if the provider is overseas.
Who regulates alternative credit scoring in Jamaica?
The Bank of Jamaica licenses and supervises credit bureaus under the Credit Reporting Act, and it also licenses microcredit lenders under the Microcredit Act 2021. A lender that builds its own scoring model from phone or bill-payment data still has to meet fair-lending and data protection rules. Talk to the Bank of Jamaica and the Office of the Information Commissioner before launch, not after.
Does place-based crime analysis work, and how is it different from predictive policing?
Place-based analysis looks at where and when incidents cluster and schedules patrols accordingly; a 2018 review by the US National Academies of Sciences found that this kind of hot-spot policing reduces crime in the short term without simply pushing it to the next street. Person-based prediction, which scores individuals on their likelihood of offending, is the version with a record of bias and wrongful targeting. A Jamaican project should stay with places and times, and publish the rules it uses.
What does a 60-day AI pilot cost a small Jamaican business?
For most service businesses the direct cost is modest: business subscriptions to general AI assistants run at roughly US$20–30 per user per month, and many accounting, hotel and customer-service systems already include AI features in higher plan tiers. The larger cost is staff time to clean data and check outputs, typically a few hours a week from one person who knows the process. Budget for that time explicitly, or the pilot stalls in week three.
Where can a Jamaican company find trained staff for AI-augmented work?
HEART/NSTA Trust runs the national training system, including programmes for the BPO sector, and can build courses with employers. UWI Mona and UTech both teach computing and data science and take industry projects. For short courses, check the current offerings of JAMPRO's sector programmes and the Business Process Industry Association of Jamaica, which coordinates training with its member companies.