I have built AI systems for Caribbean financial institutions: production systems that score real transactions and make decisions about real people's money. This piece is about what those systems do inside Jamaican banks, credit unions and insurers, and what they mean for you as a customer.
Every large financial group in Jamaica is either running AI or planning to: NCB, Scotiabank Jamaica, JMMB, Sagicor, JN Group and others. Fraud losses, customers moving to better apps, and the cost of compliance leave them little choice. What varies is how carefully they deploy it, and whether the benefits reach customers without long credit histories or good internet access.
Fraud detection: scoring each card and transfer in milliseconds
Fraud detection is the oldest AI use in banking and the one you meet most often. Each card payment, online transfer or mobile payment is scored against a model of how your account normally behaves before the merchant gets a response.
Buying groceries at Loshusan on a Saturday morning fits your pattern and goes straight through. A large purchase from a foreign website at 3 a.m., when you have never shopped abroad, gets a high score. The bank then blocks it, sends you an alert or has its fraud team call you.
Jamaica has seen more card skimming, account takeover and online payment fraud in recent years, and no bank has the staff to review every transaction by hand. The cost is false positives: a legitimate card declined at a restaurant. Banks tune their models to reduce those declines without letting more fraud through, and machine learning models make that trade-off better than fixed rules do.
Credit scoring: more data can help people with no credit file
Traditional credit scoring in Jamaica relies on credit bureau data: repayment history, current debts and length of credit history. That works for people who already have credit. Many younger adults, informal-sector workers and rural Jamaicans have little or no bureau history, so they are turned down even when they manage money well.
AI scoring can use more of the data a bank already holds: account inflows, savings behaviour, how regular income is, and whether bills are paid on time. A farmer in St Elizabeth with no credit card, but steady deposits from crop sales and a savings balance, looks creditworthy in that data.
Two risks come with this.
- Opacity. A declined applicant may not learn why, and staff may not be able to explain the model's reasons.
- Inherited bias. A model trained on past lending decisions can learn the biases in them, for example against certain parishes or occupations.
Both need rules from regulators, and banks should be able to give every declined applicant the main reasons in plain language.
Customer service: chatbots for routine questions, people for the rest
Chatbots in banking apps and websites now answer balance questions, transaction history, branch hours, rates and loan status without a wait on the phone line. NCB, Scotiabank and JMMB all use some form of this, with varying quality. The better set-ups hand complex cases to a person with the conversation attached, so you do not repeat yourself.
Inside the app, AI also drives product suggestions, unusual-activity alerts, payment reminders and spending summaries. Younger customers expect these features and will move banks for them.
Many customers, especially older ones or those making a large decision, still want to talk to a person. The banks that handle this well use AI to clear routine work so branch and call-centre staff have time for those conversations.
Anti-money laundering: fewer false alarms protects correspondent banking
Compliance matters more to Jamaica's banking system than any other AI use. Jamaican banks need correspondent relationships with foreign banks, mostly in the US, to move money abroad, and those relationships depend on strong anti-money laundering (AML) and know-your-customer (KYC) controls.
Rule-based AML systems flag transactions over a set amount, to certain countries, or matching fixed patterns. Most alerts turn out to be legitimate, and compliance staff spend their days clearing them. Machine learning systems learn which patterns are actually suspicious. Published vendor and bank case studies report false-positive reductions of 50 to 70 percent with detection held steady or improved, though results depend on data quality. For thinly staffed Jamaican compliance teams, that is time returned to real investigations.
The Bank of Jamaica has been working out its approach to AI in financial services. It has to weigh better compliance, service and inclusion against opacity, bias and privacy risk.
Insurance and investments
Sagicor, Guardian Life, JN Life and other insurers are starting to use AI in claims, underwriting and service. Models can read claim documents and photos, estimate damage and flag likely fraud. Motor claims are the highest-volume line in Jamaica, so that is where faster assessment shows up first.
JMMB, Sagicor Investments and other investment firms are adopting portfolio analysis and recommendation tools. These can give clients with small portfolios the kind of tailored advice that used to require a large account.
What changes for you as a customer
| Area | What you gain | What to watch |
|---|---|---|
| Fraud | Faster blocking of stolen-card use | Occasional wrongful declines |
| Credit | Access without a long credit file | Unexplained refusals, possible bias |
| Service | 24-hour answers to routine questions | Difficulty reaching a person |
| Data | Personalised alerts and offers | How much of your data is used, and for what |
The gains go first to customers with smartphones and reliable internet. Customers without them get less out of AI banking, and staff in routine banking roles will need new skills as the work changes.
Jamaican banking is being rebuilt on AI foundations. The banks that implement AI thoughtfully, with attention to fairness, transparency, and inclusion, will serve Jamaicans better than ever before. The ones that deploy AI carelessly or opaquely will face well-deserved scrutiny. As consumers and as a nation, we should demand the former and hold institutions accountable for the latter.
Prompts you can use with ChatGPT or Claude
Remove account numbers, TRN and passwords before you paste anything into a chatbot.
I want to understand my finances better. Here is a summary of my monthly income
and expenses in Jamaica: [details]. Help me build a realistic budget, find where
I can save, and plan to reach [goal] by [date]. Account for NHT, NIS and income tax.
I am applying for a [loan type] of J$[amount] at [bank] in Jamaica. Explain what
the bank will likely assess, list the documents I should prepare, and tell me what
would strengthen my application. My situation is: [details].
Compare Treasury bills, repurchase agreements, unit trusts, JSE stocks and fixed
deposits for a Jamaican investor. For each give the risk level, liquidity, typical
minimum investment and tax treatment. I have J$[amount] and my risk tolerance is [low/medium/high].
My loan application at [bank] was declined. Draft a polite letter asking for the
main reasons for the decision, whether an automated system was involved, whether
a person reviewed it, and what I could change before reapplying.
What to do next
- Every bank customer: turn on transaction alerts for every card and account in your banking app today, and save your bank's fraud line in your phone so you can call from the number on file.
- Anyone planning to borrow in the next year: request your free credit report from both licensed bureaus, CreditInfo Jamaica and CRIF NM, and dispute any errors before you apply.
- Declined applicants: use the prompt above to ask the lender in writing for the main reasons and whether a person reviewed the decision. Keep the reply; you need it if you escalate.
- Small business owners: run business income through one business account for at least six months before applying for credit. Consistent deposits are exactly what AI scoring models read.
- Bank risk and compliance heads: before switching on an AI credit or AML model, document which variables it uses, test approval rates by parish, age and gender, and write the plain-language reason codes staff will give declined customers.
Frequently Asked Questions
How do I get my credit report in Jamaica?
Jamaica has licensed credit bureaus under the Credit Reporting Act, including CreditInfo Jamaica and CRIF NM Credit Bureau. You are entitled to a free report from each bureau once a year, and you can request it online or in person with valid ID and your TRN. Check it for accounts you do not recognise and for debts shown as unpaid after you settled them, since lenders' models read the same data.
Who do I complain to if my bank will not fix a problem?
Start with the bank's own complaints process and ask for a reference number. If that fails, the Bank of Jamaica supervises commercial banks, merchant banks and building societies, while the Financial Services Commission supervises insurers, securities dealers and pension plans. Send the regulator your reference number and copies of your correspondence.
Does the Data Protection Act give me any rights over automated credit decisions?
Yes. Jamaica's Data Protection Act 2020, in force since December 2023, includes rights over decisions based solely on automated processing, and gives you the right to ask a bank what personal data it holds about you. Ask whether a person took part in the decision and request a review if not. The Office of the Information Commissioner handles complaints under the Act.
What is the Bank of Jamaica's fintech sandbox?
It is a regulatory sandbox the Bank of Jamaica opened in 2020 that lets fintech firms test new payment and financial products with real customers under supervision and set limits. Products that pass can then seek a full licence. For customers, a product being in the sandbox means it is supervised but still in a testing phase, so check the limits on what it can hold.
Why did my card get declined when I was not doing anything unusual?
Fraud models score each transaction against your history and against current fraud patterns across the bank. A merchant or location that has recently seen skimming can raise your score even if your own behaviour is normal. Call the number on the back of your card, confirm the transaction, and tell the bank about planned overseas trips or large purchases in advance.