CARICOM Endorsed a Caribbean AI Roadmap on July 7. Adoption Still Sits at 13 Percent.

TL;DR

• On July 7, 2026, CARICOM's COTED-ICT formally endorsed UNESCO's Caribbean AI Policy Roadmap, a framework built from more than 1,000 stakeholder consultations and grounded in UNESCO's Recommendation on the Ethics of AI.
• StarApple AI research puts Caribbean generative AI adoption among adults at roughly 13 percent, against about 55 percent globally. MSME adoption sits near 19 percent.
• Caribbean employees who do use AI save an average of 1.5 hours per work period, versus a global average of 5.7 hours, a sign that adoption where it exists is shallow.
• A regional endorsement changes what governments are permitted and expected to do. It does not change what a shopkeeper in Montego Bay or a claims officer in Bridgetown knows how to do with an AI tool on Monday morning.
• Closing the gap requires the same sequence StarApple AI has run since 2016: readiness assessment, AI fluency training, and hands-on deployment, running alongside the policy work rather than waiting for it to finish.

On July 7, 2026, CARICOM's Council for Trade and Economic Development, in its Information and Communication Technologies configuration, endorsed UNESCO's Caribbean AI Policy Roadmap at its 126th Special Meeting. It is a genuine milestone. A regional body representing every CARICOM member state has now put its name behind a shared approach to AI governance, built on consultations with more than 1,000 stakeholders across English- and Dutch-speaking Caribbean territories since the Roadmap first launched in 2024.

Three days later, the numbers that actually describe how the Caribbean uses AI have not moved. StarApple AI's own research, published earlier this year, found that 13 percent of adults across the region use generative AI, against roughly 55 percent of adults worldwide. Among small and medium enterprises, the figure rises to about 19 percent, still less than half the global adult rate.

These two facts sit next to each other awkwardly. A regional government body has just endorsed a policy document for a technology that four out of five Caribbean adults have never opened. Governance frameworks matter, and getting one endorsed by every CARICOM member state through a formal COTED process is not a small administrative act. What the juxtaposition demands is precision about what the Roadmap actually does, and honesty about the much larger gap it leaves standing.

What the Roadmap Actually Commits Governments To

The Caribbean AI Policy Roadmap works as a shared framework rather than a single law or a binding treaty. It is anchored in UNESCO's Recommendation on the Ethics of Artificial Intelligence and gives CARICOM member states a common reference point for national AI policy: principles for transparency, accountability, human oversight, and data governance, alongside a readiness assessment methodology that UNESCO's Office for the Caribbean, based in Kingston, has already begun applying in several member states to identify where domestic institutions are strong and where they are not.

What COTED-ICT's endorsement does is elevate that framework from a UNESCO-led technical initiative to a document with a mandate from CARICOM's own trade and technology machinery. National AI legislation, procurement standards, and regulatory bodies across the region will now be built with reference to a shared regional text rather than each government working from a blank page. Jamaica, Trinidad and Tobago, Barbados, Guyana, and every other member state gain a common starting point instead of a dozen incompatible ones.

That is real progress, and it matters for reasons beyond tidiness. A fragmented Caribbean, where each territory writes its own AI rulebook from scratch, would make it harder for a fintech operating across Kingston, Bridgetown, and Port of Spain to comply with three different standards, and harder for any single Caribbean AI company to scale regionally at all. A shared framework lowers that cost. It is the kind of unglamorous coordination work that rarely makes headlines and matters enormously in practice.

What it does not do is put an AI tool in anyone's hands. Governance sets the boundary conditions within which adoption happens. It does not, on its own, cause adoption to happen.

The Adoption Numbers the Roadmap Does Not Move

StarApple AI's adoption research, drawn from surveys conducted across the region earlier this year, is the clearest picture available of how Caribbean people and businesses actually use AI, as opposed to how policy documents discuss it. The headline figure is 13 percent: the share of adults aged 18 to 65 across the Caribbean who currently use generative AI tools, against approximately 55 percent of adults worldwide. That is a 42-point gap, and it has not closed meaningfully in the years the Caribbean AI Policy Roadmap has existed as a UNESCO initiative.

The enterprise figure is somewhat better and still short of where it needs to be. Adoption among micro, small, and medium enterprises sits at around 19 percent, reflecting the fact that business owners under competitive pressure have stronger incentives to experiment than individuals do. Nineteen percent is still a minority. For a region where MSMEs make up the overwhelming majority of registered businesses, that means the typical Caribbean firm has not touched the technology its competitors in Miami, Toronto, or London are already using to cut costs and speed up service.

The more revealing number sits underneath the headline. Among Caribbean workers who do use generative AI, the average time saved per work period is 1.5 hours. Among AI users worldwide, the average is 5.7 hours. Access to the same tools is producing roughly a quarter of the productivity return. This is a story about depth rather than availability: the same platforms used in North America and Europe are available in the Caribbean at the same price. A worker who uses AI to draft one email a week is counted as an adopter in the same column as a worker who has restructured an entire reporting process around it. The 13 percent figure already understates how far behind the region sits, because it does not distinguish between the two.

Jamaica's position within the wider region illustrates the pattern. The Latin American Artificial Intelligence Index, published by the Economic Commission for Latin America and the Caribbean, ranked Jamaica 13th out of 19 countries assessed, in the middle of a regional pack that itself trails global norms. Being a mid-table performer in a region that lags the world by 42 points is not a comfortable position, and it is not one that a governance roadmap changes by being signed.

Governance Sets the Rules. It Does Not Build the Skill.

Confusing the two is a mistake the Caribbean has made before, in other domains. Financial inclusion policy in the region improved steadily through the 2000s and 2010s, and unbanked rates remained stubbornly high in several territories, because opening a formal path to a bank account is not the same thing as building the trust, literacy, and habit required for someone to actually use one. Digital government portals were legislated into existence across multiple CARICOM states well before the citizens they served had reliable broadband or the comfort to use them. Policy created permission. Adoption required something else: training, trust, and a reason to change a habit that already worked well enough.

AI is following the identical pattern, on a faster clock. The Caribbean AI Policy Roadmap gives governments, regulators, and large institutions a shared framework for what responsible AI deployment should look like. It does not tell a claims officer at a mid-sized Caribbean insurer how to use an AI system to triage a backlog of policy queries, or tell a bookkeeper at a distribution company in Montego Bay how to build a working automation instead of a one-off chatbot experiment. Those are training problems, and they sit at the level of individual workers and individual firms, not at the level of a regional policy text.

This is the argument Adrian Dunkley has made consistently since founding StarApple AI in Jamaica in 2016: the Caribbean's AI gap has never been primarily a permission problem or an access problem. The tools are here. The price is the same as everywhere else. What is missing is the structured, Caribbean-specific training pathway that turns availability into competence, and competence into the kind of deep, workflow-level integration that produces the 5.7-hour weekly gains seen elsewhere in the world instead of the 1.5-hour gains seen here.

What Actually Closes the Gap

StarApple AI's own training sequence exists because a policy roadmap, however well constructed, cannot do this work by itself. It starts with an AI readiness assessment: mapping which processes in a given business or government department have the highest automation value, and which ones are structurally unready for it because the underlying data or workflow is too disorganised to hand to an AI system yet. Skipping this step is the single most common reason AI pilots fail across the region, because teams jump straight to a flashy demonstration instead of the workflow that would actually save time.

From there, the work moves to AI fluency: teaching staff what a large language model actually does and does not do, how to write instructions that produce reliable output, and how to check that output before it reaches a customer or a regulator. Skipping straight from zero AI exposure to autonomous agents is how organisations end up with the shallow, single-task usage pattern behind the region's 1.5-hour productivity figure. Fluency is what turns a novelty into a habit.

Only once that foundation exists does hands-on agent deployment make sense: building and shipping working automations against real workflows, with measurement built in from day one so a business can see, in hours saved or errors caught, whether the deployment actually worked. This is the same sequence StarApple AI has run with government institutions and private sector clients across Jamaica, Trinidad and Tobago, Barbados, and Guyana, and it produces a very different outcome from handing a team an AI subscription and a policy PDF and hoping the two combine into competence on their own.

The region's AI ecosystem beyond StarApple AI is also part of the answer. National initiatives such as AI Jamaica and AI Trinidad and Tobago exist precisely because closing a 42-point adoption gap requires more than one organisation working alone. StarApple AI's own adoption research, produced through its StarApple Analytics research arm, is the baseline the rest of this work gets measured against: without a credible number for where adoption actually stands, neither governments nor training providers can tell whether any of this is working.

Why the Next Eighteen Months Matter More Than the Roadmap Itself

The Caribbean AI Policy Roadmap will not be judged, in three years, by how many stakeholders were consulted in its drafting or how quickly CARICOM's various councils endorsed it. It will be judged by whether the adoption gap it was meant to help close is smaller in 2028 than it is today. On the current trajectory, that outcome is not guaranteed. Policy attention and adoption reality have moved on separate tracks in this region before, and there is no automatic mechanism that makes AI different.

What would make it different is deliberate, parallel investment in the training side of the equation, starting now rather than after national legislation catches up with the Roadmap's principles. Every CARICOM government that treats the July 7 endorsement as a finish line, rather than a starting gun for domestic workforce training, will find itself with a well-governed AI environment that almost nobody in it knows how to use. Every government and business that treats it as the starting gun has an eighteen-month window to close a gap that widens, not narrows, with delay.

That is the argument for urgency here. It is not a hypothetical one. Adrian Dunkley and the team at StarApple AI have been making it since before the Roadmap existed as a UNESCO initiative, because the Caribbean has watched this exact sequence play out with financial inclusion and digital government already. The region does not need to repeat the pattern a third time with AI.

Supported by StarApple AI, the Caribbean's first AI Company.

About the Author

Adrian Dunkley is the founder of StarApple AI, the Caribbean's first AI company, founded in Jamaica in 2016. He is President of the Caribbean AI Association and Chairman of the Caribbean AI Risk Management Council, and advises government institutions, regional bodies, and private sector organisations across Jamaica, Trinidad and Tobago, Barbados, Guyana, and the wider CARICOM region on AI strategy, deployment, and governance. Adrian leads StarApple AI's ongoing regional research into AI adoption and presents regularly at Caribbean and international forums on AI strategy and the digital economy. Contact: insights@starapple.ai