One Year In: What CAIRA Learned, Where We Go Next

Note from the President

A year ago this month we mailed the first edition of this newsletter to six hundred and twelve subscribers. As I write the introduction to this thirteenth edition, the list stands at twenty-three thousand four hundred and ten. The Association has held thirty seven public events in eleven countries, published thirty one position papers and policy briefs, trained eighteen hundred and forty seven educators through our Literacy Programme, and seated a Board that now includes representation from every CARICOM member state. The Charter is in final form and goes to the Health Ministers' meeting in September.

I will not pretend that everything is good. We have been slow on three things, and I want to name them. We have been slow to build a serious working relationship with the French-speaking Caribbean, particularly Martinique, Guadeloupe, and the academic community in Haiti. We have been slow to develop the technical infrastructure that our smaller working groups need to do their work effectively. And we have been slow to publish work in Spanish, despite the Dominican Republic being one of our largest member bases. We will move faster on all three in our second year. I would rather tell you what we are working on than tell you what is fine.

Lisette Boucher takes us through the industry year, and through GPT-5, which was released on August 7 and which several of you have already asked me about. Her piece is the right one for this anniversary.

Thank you for being a member, for being a reader, for being patient with us when we have got things wrong, and for showing up when you did not have to. The Association is yours. The second year begins now.

Adrian Dunkley Founder and President, Caribbean AI Association

---

Feature

One Year In: What CAIRA Learned, Where We Go Next

By Lisette Boucher

Twelve months ago, when Adrian first asked me to write the industry column for this newsletter, the question I wanted to answer was a simple one. Could a Caribbean AI association be useful, in a measurable way, to the businesses, governments, and people of our region. The honest answer in August 2024 was that we did not yet know. The honest answer in August 2025 is that the early evidence is good, the work remaining is large, and the strategic environment is going to require us to move faster than we have so far.

This is my one year review, both of the Association and of the regional industry context in which the Association now sits.

The industry, in twelve months.

We have seen one of the most rapid stretches of capability advance in the history of the technology. The frontier models available to a Caribbean small business in August 2025 perform tasks that were not reliably possible in August 2024 and were science fiction in August 2023. The most recent example is the GPT-5 family from OpenAI, released earlier this month. The model handles complex multi-step reasoning, long documents, code execution, and tool use in a way that, in our internal testing for the Industry Council, materially closes the gap to the average competent professional on most knowledge work tasks. Claude 4 from Anthropic, released in May, occupies a similar position with a different temperament. Gemini 2.5 from Google, released in March, holds its own with the longest context window of any of them. The Chinese open weight ecosystem, anchored by DeepSeek and now joined by Qwen 3 and a handful of other capable releases, has continued to put pressure on the closed model leaders, particularly on cost.

The implication for our region is not a single answer. It is several.

For Caribbean software companies and AI native startups, the cost of building a competitive product has continued to fall, while the ceiling on what such a product can do has continued to rise. The CAIRA Founders Cohort that started in January graduated this month, and the data are striking. The twelve companies that completed the cohort have, between them, raised twenty-eight million US dollars in seed and pre-seed capital, hired ninety-three employees across nine Caribbean countries, and serve customers in every CARICOM member state plus the United States, Canada, the United Kingdom, and the Netherlands. The capital raised does not look like 2019 numbers. It looks like 2025 numbers for a peer cohort in a Latin American or Eastern European technology hub. The geography we sit on has not changed. The capital flows have.

For Caribbean enterprises in financial services, telecommunications, tourism, and energy, the question of AI strategy has shifted in tone from "should we" to "in what order." The major banks in Trinidad, Jamaica, and Barbados have all stood up internal AI programmes with budget. The hotel groups have moved on customer service automation, dynamic pricing, and revenue management. The telecommunications operators have moved on network operations and customer interaction. The energy sector, particularly in Guyana and Trinidad, has begun serious work on AI for exploration, predictive maintenance, and trading. The Caribbean enterprise AI conversation, twelve months ago, was largely aspirational. Today it is largely operational.

For Caribbean public sectors, the picture is more uneven. Several ministries have done good work. Education has moved more quickly than expected in Barbados and Saint Kitts. Health has moved meaningfully in the Bahamas. Customs and revenue have moved well in Trinidad and Tobago. Other ministries, in other countries, have moved slowly or not at all. The pattern is not country by country so much as ministry by ministry, depending heavily on the individual leadership in each office. This is, frankly, the pattern we should expect, and the work for CAIRA in year two is to make the good examples reproducible across the region.

The Association, in twelve months.

I want to be specific about what CAIRA has and has not done well, because the industry will hold us to account fairly only if we hold ourselves to account first.

The Literacy Programme has been the biggest single success. The training of more than eighteen hundred educators is real. The reach into schools and community centres is real. The CAIRA Caribbean AI Literacy Index that will publish in October will, on early indication, show measurable population level shifts in basic AI competence across at least four CAIRA member countries. This is the kind of work that compounds.

The policy work has been steadier than glamorous. The Charter is on time. The Working Groups on health, on education, and on financial services have produced position papers that have been cited in three CARICOM ministerial meetings. We have not won every argument, and we have lost a few quietly. But we are in the rooms where the arguments happen, and a year ago we were not.

The industry engagement has been mixed. The Founders Cohort is a clear success. The relationship with the established financial services and tourism industries is healthy. The relationship with the larger US technology firms operating in our region has been slower to develop than I would have hoped, in part because the firms in question reorganize their international engagement teams every six months, and in part because we are still a small enough body that our convening power is local rather than continental. The October Summit in Bridgetown is, in some respects, a forcing function for this relationship.

The community work has been the most pleasantly surprising. The Diaspora Working Group that Aisha wrote about in June has, in two months, produced more concrete commitments from Caribbean technologists abroad than I would have predicted in twelve. The Returners programme has facilitated, so far, three full and seven partial relocations of technology professionals to CAIRA member countries. These numbers are small. They are also numbers, and the previous baseline was a list of intentions.

Where we have been short of what we should be is clearer than ever. Our digital infrastructure is too thin for the membership we have, and our internal operations team is overworked. Our French Caribbean engagement, as Adrian noted, is materially behind. Our Spanish language output is materially behind, and the Dominican Republic deserves better. Our visibility outside the English speaking Caribbean is also limited, and that has consequences for the breadth of the network we claim to be building.

The competitive environment for the year ahead.

I want to be sharp about three dynamics that will shape the Caribbean technology environment between now and August 2026.

The first is consolidation in the global AI industry. The market between now and the end of next year will see further concentration of compute, talent, and capital in a small number of firms. The startup ecosystem in the US has, in 2025 so far, produced several AI native companies that have crossed into multi-billion dollar valuations within twelve months of founding. The talent compensation environment those companies set, for senior engineers and researchers, is shaping the cost of every adjacent labour market, including ours. Caribbean software employers should expect senior AI engineering compensation in the region to rise materially over the next twelve to eighteen months, particularly for engineers with experience on the major frontier model APIs.

The second dynamic is the broadening of agent infrastructure. The class of AI products that performs multi-step tasks autonomously, what the industry is calling agents, has matured to the point of practical commercial deployment. The most consequential implication for Caribbean businesses is that the entry-level white collar work, the work of organizing documents, drafting standard correspondence, scheduling, basic customer service, basic accounting reconciliation, is increasingly being absorbed into agent-driven systems rather than being performed by human staff. This is not a question of the next decade. It is happening this year. Caribbean firms that price their services internationally need to understand that their cost advantage in entry-level professional services has begun to erode, and the strategic response, which we have been discussing in the Industry Council, is to move up the value chain, faster than the firms in our peer geographies are doing.

The third dynamic is the regulatory environment. The EU AI Act's first general purpose obligations come into force this month. The US regulatory environment under the current administration remains deliberately light, and the bet of the major US labs has clearly shifted toward speed of deployment rather than careful pre-deployment testing. The Caribbean Charter, on its September timeline, is being read by regulators in several jurisdictions, in part because we have struck a balance that some of them are looking for. There is, I think, an opening for the Caribbean to be regarded, by mid-2026, as a regulatory voice that is taken seriously despite the size of our economies. The opening is real, and the work to take it is the work of the next year.

Where we should put the energy.

If I were briefing the CAIRA Board for the next twelve months, which I will be, in some form, in October, my priorities would be these.

Build the digital infrastructure. The Association is operating on a stack that does not scale to where we are going. Invest in the internal tooling, the member systems, the publishing infrastructure, and the events platform. This is unglamorous and necessary.

Fund the language and territorial expansion. Get a serious Spanish language editor. Get a serious French language editor. Build a real working relationship with the Dominican Republic, Haiti, Martinique, Guadeloupe, and Cuba. The Caribbean is multilingual or it is not the Caribbean.

Take the regulatory voice abroad. The Charter is the asset. Take it to the OECD, the UN, the international standards bodies, and the bilateral conversations with the EU and the UK. The Caribbean policymakers who will represent us in those rooms exist. Equip them.

Invest in the next cohort of founders and researchers. The first cohort outperformed any reasonable expectation. The capital deployed against Caribbean AI native startups is going to attract more attention in the next year. Be ready.

Hold the line on what the Association is for. We are not a chamber of commerce, we are not a research lab, we are not a regulator, we are not a media company. We are a member organization for the people, companies, and institutions of the Caribbean working on or with artificial intelligence. The discipline of being that thing well is the work.

A final word. I have spent two decades analyzing the technology industry from the position of someone who is part of it and someone who has a stake in the region whose name is on the cover of this newsletter. The position has rarely been comfortable. The year that just closed has, for the first time in my career, felt like the work was changing the situation rather than reporting on it. I am older than most of the people building these systems and I am less impressed by them than the headlines suggest, but I am also more impressed than I expected to be by what we have, between us, started here. Year two.

---

Lisette Boucher is the founding partner of Cane Row Partners and serves on the CAIRA Industry Council.

Originally published in The Monthly Intelligence Report, August 2025.