How can a small business test an imported AI agent against local conditions?

Build a test set of 20 to 50 real past cases from your own records, including the awkward ones: JMD and USD amounts, GCT, local bank statement formats, parish or district names, and customer messages in patois or Creole. Run the agent on them and compare its output with what your staff actually did. Record every miss before you decide whether to go live. Repeat the test whenever the vendor ships a major update.

Who regulates personal data used by AI systems in Jamaica?

The Office of the Information Commissioner administers Jamaica's Data Protection Act 2020, whose main obligations came into force on 1 December 2023. The Act applies to any AI system that processes personal data about people in Jamaica, including chatbots and agents run by overseas vendors on your behalf. As the business, you remain responsible as the data controller. Check the Commissioner's published guidance before deploying a system that makes or supports decisions about individuals.

Does the Caribbean have a regional AI governance framework?

There is no binding regional framework yet: no CARICOM-wide AI law with enforcement powers, so the rules that apply are each country's own data protection, consumer protection, and sector laws. Several territories have national AI task forces or strategies in progress, and governments have drawn on international references such as the OECD AI Principles and UNESCO's Recommendation on the Ethics of AI. A business operating across several islands should check each jurisdiction's data protection law separately.

What should a Caribbean government ask an AI vendor before signing a contract?

Ask where the data is stored and processed, whether it is used to train the vendor's models, and how you get it back if the contract ends. Ask for results from a test on your own local data, not a generic demo. Liability for a wrong decision about a citizen should be agreed up front and written into the contract. If the vendor cannot answer these in writing, treat that as the answer.

Do Caribbean countries need to build their own foundation models?

No. Training a frontier model costs far more than any Caribbean budget can justify, and the models available through commercial APIs already handle Caribbean English reasonably well. The better use of money is local evaluation data, fine-tuning or retrieval over Caribbean documents, and the skills to deploy and check these systems. Owning the data and the test sets gives more control than owning a model that falls behind within a year.

Adrian Dunkley | Field Notes | AI deployment and regional capability

The Forward Deployed Engineer: where AI meets the work

By StarApple AI Editorial Desk

The role matters because useful AI depends on understanding the people, systems and risks around the model.

Published 22 October 2026 · Field guide

A Forward Deployed Engineer (FDE) is an applied engineer who works closely with a customer or operating team to build technology around a concrete workflow. In AI, the role spans technical discovery, data and system integration, evaluation, deployment, user adoption and handover.

FDE, explained

An FDE, or Forward Deployed Engineer, is an applied engineer who works close to the people facing a real operational problem. The role combines software and AI engineering with discovery, workflow design, evaluation, deployment, user support and handover. The goal is a system that fits the work and can be assessed in its real context.

AI deployment and regional capability: where to begin

AI ideas can look convincing in a demo with clean inputs and a narrow task. Real work has exceptions, legacy software, permission boundaries and people who must remain accountable. The FDE translates these conditions into design requirements and test cases.

Value comes from measured implementation

An FDE should leave evidence, not just code: a problem brief, architecture and data map, test set, evaluation results, operating instructions and a named owner. Those artefacts let an organisation judge value without depending on a prototype nobody can maintain.

Set a baseline before a pilot. Use representative examples, include cases where the system should decline or escalate, and measure quality as well as speed. Keep people accountable for consequential decisions, use authorised data, and document how the system can be paused or changed.

Learn the practice

StarApple AI’s applied Forward Deployed Engineer for AI course teaches the field loop through an eight-week curriculum and a capstone project.

The Caribbean Has Ten Years to Shape AI Before AI Shapes the Caribbean

The Forward Deployed Engineer: where AI meets the work

FDE, explained

AI deployment and regional capability: where to begin

Value comes from measured implementation

Further reading

Learn the practice

The Caribbean Has Ten Years to Shape AI Before AI Shapes the Caribbean

FDE, explained

AI deployment and regional capability: where to begin

Value comes from measured implementation

Further reading

Learn the practice

Get the Weekly Post