When Caribbean people show up in Anthropic's data at all, they are mostly developers writing code, marketing teams writing copy and staff drafting customer service replies. That data is Anthropic's Economic Index, the report built from anonymised Claude conversations, which I call the Claude Economic Index here. It barely registers the Caribbean, and where our countries appear, usage is low. Policymakers should read that as a measurement of where we are, with some blind spots I explain below.
I have built AI systems in Jamaica since 2011, founded StarApple AI in 2016, and run four AI labs here. This piece covers what the index measures, where the Caribbean sits in it, what it cannot see, and what governments should do with the result.
The index counts what people do with Claude, not what they say they do
Anthropic analyses samples of Claude conversations with a privacy-preserving tool, strips out identifying details, and classifies each conversation by task and occupation. It then compares countries by usage per working-age person, so a small country is not penalised for being small.
That is its strength. Most studies of AI adoption rely on surveys, where companies report what they plan or hope. The index reports observed behaviour: which tasks people bring to an AI assistant, and whether they use it to do the work for them or to help them do it.
Caribbean usage is low and stuck on coding, copy and customer replies
Measured by usage per person, the Caribbean sits below Latin American peers such as Brazil, Mexico and Colombia, and far below small, high-adoption economies such as Singapore and Estonia. Anthropic's own analysis finds usage rises with income per head, which explains part of the gap but not all of it.
The Caribbean usage that does appear falls into three familiar buckets: developers writing code, marketing teams writing copy, and businesses drafting customer service replies. Those are real uses. They help a person do an existing task faster and leave the business model unchanged.
The index shows little Caribbean use in credit modelling, crop planning, public service delivery or clinical support. The same report found that low-adoption countries lean heavily on coding while high-adoption countries spread their use across many more kinds of work, and the Caribbean looks like the former.
Our three biggest earners are being retooled elsewhere
Tourism, remittances and commodity exports carry most Caribbean economies, and competitors in each are already using AI in their operations:
- Hotel groups in Southeast Asia and elsewhere use dynamic pricing and personalised offers, competing for the same long-haul guests as Negril and Grace Bay.
- Remittance providers use automated identity checks and fraud screening to cut the cost per transfer, which pressures local money transfer agents.
- Agricultural exporters in South America and Africa use computer vision for grading and traceability, which buyers in the UK and US increasingly expect.
A hotel that uses AI only to write its Instagram captions is not competing on the same terms as one that prices every room nightly with a model.
What the index cannot see in the Caribbean
The index has four blind spots that matter here, and they cut both ways.
- It measures one product. Many Jamaican professionals I work with use ChatGPT or Gemini daily, and none of that appears in Anthropic's data.
- It misses custom systems. Fraud models at Caribbean banks, credit scoring at credit unions and the systems we build at StarApple AI run outside commercial assistants. Their effect shows up in fraud losses and loan approvals.
- It does not adjust for the cost of connectivity. The UN Broadband Commission's affordability target is entry-level broadband at under 2% of monthly income per head. Where a household pays more than that, lower usage partly reflects price rather than interest.
- Small samples are noisy. A country with a few hundred thousand adults generates few conversations, so year-on-year changes for individual Caribbean islands should be read with caution.
Even allowing for all four, I do not think the gap disappears. The custom systems I know of amount to a handful of projects.
Where the Caribbean should aim first
I built the LATAMC AI Playbook, which maps more than 15,000 AI use cases across Caribbean and Latin American sectors, so that governments and firms do not have to learn by trial and error. Set against the index, it points to four areas where the return is highest for us: financial inclusion, tourism personalisation, agricultural traceability and automation of routine government services. Each has working examples in other markets with similar constraints.
What Caribbean governments should do next
- Statistical institutes (STATIN, CSO Trinidad and Tobago, Barbados Statistical Service): add three questions on AI use to the next labour force or business establishment survey. Asking which tools, for which tasks, and how many staff use them each week gives a baseline the index cannot provide.
- Ministries of finance: put AI infrastructure (data sharing between agencies, cloud budgets, staff training) in the capital budget as a line item with a figure. A declaration signed at a conference buys nothing.
- Cabinet offices: set dated targets. Singapore's National AI Strategy 2.0 aims to reach 15,000 AI practitioners. A Caribbean equivalent reads like "AI triage support in half of public health centres by 2028", with a named ministry accountable.
- Permanent secretaries: before building in-house AI teams, contract local firms and university labs for the first two pilots. The expertise in the region sits mostly in the private sector and at UWI, and a pilot run with them is faster than a new unit hired from scratch.
- Regulators and ministries of science and technology: review the index's country data each time Anthropic updates it, alongside ChatGPT and Google usage reports, and publish a short regional note so that one company's data does not become the only picture of us.
What I cannot tell you from this index is how much of the gap is price and how much is demand. Only a local survey will answer that, which is why the first action above comes first.
Frequently Asked Questions
Where can I see the Claude Economic Index data for my own country?
Anthropic publishes the reports and an interactive explorer on its website under Economic Index, and releases the underlying dataset on Hugging Face under the Anthropic organisation. Search the country-level file for your country code. If your island is missing, it usually means the sample was too small to report reliably, not that usage was zero.
Does the index read people's private conversations?
Anthropic says it uses an automated, privacy-preserving analysis system that summarises conversations into aggregate categories, and researchers see only those aggregates. Country figures are reported only above a minimum number of users. If you want to keep a conversation out of this kind of analysis entirely, read Anthropic's current privacy policy and your account's data settings.
What is the difference between automation and augmentation in the index?
Augmentation is when a person works with the AI, for example asking it to check or improve their draft. Automation is when the person hands over the task and takes the output with little further input. The balance matters for jobs: a country moving towards automation is handing over whole tasks.
Is there a Caribbean source of AI adoption data besides the index?
Not yet a regular one. The Caribbean Development Bank, the IDB and ECLAC publish occasional digital readiness studies, and UNESCO's AI Readiness Assessment Methodology gives governments a template for measuring their own position. National statistical institutes are best placed to collect a recurring figure through their existing surveys.
Can a Caribbean business use the index to plan its own AI projects?
Yes, as a list of tasks others already bring to AI. Look up the occupations in your company in the index's task breakdown, note which tasks show heavy AI use elsewhere, and test those first. It will not tell you the return, so measure time saved on a small pilot before rolling anything out.
"The Claude Economic Index measures which economies are putting AI to work and which are watching. The Caribbean has to decide which side of that line it wants to be on." - Adrian Dunkley, AI Boss