The Caribbean AI Adoption Imperative: Why Caribbean Businesses Cannot Sit This One Out

 

IN THIS ARTICLE This is the first of twelve articles. It argues that a technology cycle arriving now will reshape Caribbean enterprise faster than any before it — and that the time available to act is shorter than most boards are treating it as. This series will equip Caribbean executives and board members with a structured, evidence-based understanding of artificial intelligence adoption in our region. This opening article sets out why the cycle matters, how it differs from previous technology waves, and what a credible executive response looks like today. By the end of this article you will be able to: 1.   Explain to your board, in plain language, why the current AI cycle is structurally different from the cloud migration, the mobile revolution, or the blockchain hype cycle — and why that difference matters for the Caribbean specifically. 2.   Recognise the four compounding factors that will make late AI adoption meaningfully more expensive than timely adoption — and use them to assess your organisation’s current position. 3.   Apply the D-AGENTICA™ Three-Question Board Diagnostic to surface, in a single leadership meeting, where your organisation actually stands on artificial intelligence — and where it needs to go.

A board chair I know well, one of the most thoughtful non-executive directors in the Caribbean financial services sector, recently concluded a board meeting by asking his chief executive what the institution’s strategy was on artificial intelligence. The CEO, a capable executive with more than two decades of operating experience, paused for a moment and then said something that has stayed with me. He said, ‘We are watching it. We have not seen a compelling reason to move yet.’

The chair nodded, moved to the next item on the agenda, and the meeting closed. In the car afterwards, the chair telephoned me. He said, ‘Dawkins, I am not sure what a compelling reason would look like to him, and I am not sure what a compelling reason should look like to me. Help me think about this.’

I have had some version of that conversation seven times in the last ninety days. Different territories, different sectors, different institutions — banks, insurers, credit unions, a hospitality group, a regional manufacturer, a public-sector chief executive, a law firm managing partner. In each case the same structure: a senior, capable person who reads widely and who is aware that artificial intelligence is a topic they should have a position on, but who has not yet developed one they feel they could defend. And in each case the same underlying question, whether they phrased it this way or not: how do I tell a real shift from a hype cycle, and if this one is real, what am I supposed to do about it?

This series is my attempt to answer that question for Caribbean executives and board members. Over the coming three months, across twelve articles, I will lay out what I believe the evidence says, what the Caribbean context adds to it, what the realistic response is for organisations of the sizes and shapes we actually have in this region, and what the cost of delay is likely to be if we get the timing wrong.

I want to be honest with you at the outset about my own perspective. I have spent the last three years watching artificial intelligence arrive in Caribbean boardrooms, and the last eighteen months advising institutions that have been trying to navigate it. I am not a technologist. I am the executive chairman of a professional services firm that has been built around audit discipline, regulatory fluency, and the specific realities of Caribbean enterprises. What I bring to this conversation is not Silicon Valley enthusiasm. It is the same disciplined scepticism that my profession brings to every significant business decision, applied to a topic that most of my peers are still circling at a distance.

My argument, laid out across the rest of this article, is that the current technology cycle is genuinely different from the ones that came before it, that the difference has specific consequences for Caribbean enterprises, and that the window for timely action is meaningfully shorter than Caribbean boards are currently treating it as. I will explain what has changed in the technology itself, what the productivity evidence looks like from sources that are not selling anything, and what the specific Caribbean dynamics of this cycle are. At the end of the article, I will give you three questions you can take to your own leadership team this quarter that will tell you, in a single meeting, where your organisation actually stands.

By the time you finish this article, you will not have a complete AI strategy. But you will have the beginning of one — enough to ask better questions, to challenge the vendors who are already calling on your organisation, and to decide whether and where to move first.

What has actually changed in the technology

Every technology cycle produces a chorus of voices claiming that this one is different. Most of the time, those voices are wrong. The claim has been made about blockchain, about the metaverse, about the Internet of Things, about 3D printing at industrial scale. I have made some of those claims myself over the course of my career, and I have later had to explain to clients why the payback never arrived. So when I tell you that the current cycle is genuinely different, I am asking you to trust that I have applied the same scepticism to this claim that I apply to all the others.

Here is what is different, stated as precisely as I can.

The previous generation of artificial intelligence — the generation that gave us the chatbots, the recommendation engines, the image classifiers that have populated enterprise software since roughly 2017 — operated under a specific limitation. Those systems responded to a prompt, produced an output, and stopped. They were functions, in the computational sense of that word. You asked them something, and they answered. The value they created was real, but the shape of that value was confined to the edges of workflows. The human was still doing the work. The AI was helping with discrete steps along the way.

The systems arriving now do not stop. When you give one of them a goal — produce a reconciliation, draft a response to a customer, review a lease agreement, identify variances in a set of quarterly results, prepare a working paper for a tax filing — the system decomposes the goal into sub-tasks, retrieves the information it needs, calls the software tools it has been given access to, produces intermediate work, checks its own output against the original goal, and returns a finished product. When something goes wrong along the way, it notices, and it tries a different approach. When it reaches a decision it is not authorised to make, it escalates to a human being. Crucially, it does not need a human in the loop at every step. It needs a human at the boundaries of the work.

The industry has settled on a single word for these systems. The word is agents. It is a clumsy word, because it invites confusion with actuarial agents and travel agents and the unrelated banking product named Agent IQ. But the word we have is the word we have, and it captures something real. An artificial intelligence agent is a software system that can plan, use tools, and execute multi-step work against a goal. It is not a smarter chatbot. It is a different category of thing.

 

An AI agent is not a smarter chatbot. It is a different category of thing.

The practical consequence of that distinction is what matters. The work these systems are capable of doing is no longer confined to the edges of workflows. It reaches into the centre. Anything that can be described as a procedure — which is to say, most of the work that most Caribbean enterprises actually pay people to do — is now within the reach of agentic automation. This does not mean automation in the sense of replacement. It means automation in the sense of augmentation — of a single competent professional being able to do the work that previously required three, or four, or in some functions ten.

I want to be careful here, because this is where Caribbean executives are most likely to be misled, and I do not want to contribute to the misleading. The claim I am making is not that these systems can replace professional judgment, replicate decades of institutional knowledge, or operate safely without human oversight. They cannot, and any vendor selling you a system that claims otherwise is selling you a reason to be sued. The claim I am making is much more specific. It is that the routine execution layer of professional work — the preparation, the search, the cross-referencing, the drafting, the calculation, the reconciliation, the reporting — is being compressed by a factor that most Caribbean executives would find implausible if I quoted it without evidence.

So let me quote the evidence.

What the productivity evidence actually says

I am going to restrict myself, in this section, to evidence that comes from peer-reviewed academic studies, from reports issued by organisations with reputations to protect, and from the audited disclosures of publicly listed companies. I am not going to rely on vendor case studies, which are unreliable in every technology cycle and particularly unreliable in this one, where the commercial incentive to overstate productivity claims is enormous.

On software engineering, a randomised controlled trial published in 2023 by researchers at GitHub — the code-hosting platform — found that developers working with an AI coding assistant completed a defined programming task 55 percent faster than developers not using the assistant. The trial controlled for developer experience and for task difficulty. Code quality, measured independently, was not statistically different between the two groups. This was measured on a first-generation assistant. The systems available to enterprise buyers in 2026 are materially more capable than the one studied.

On customer service, a study of more than 5,000 customer support agents working in a large North American contact centre, conducted by researchers at Stanford and the Massachusetts Institute of Technology and published in 2023, found that AI-augmented agents resolved 14 percent more issues per hour on average than their unaugmented colleagues. The study is particularly interesting because the largest productivity gains appeared among less experienced agents. AI appears to lift the bottom of the distribution faster than it lifts the top. The study was conducted before the current generation of agentic tooling was commercially available. Recent deployments of agentic systems in financial services contact centres are reporting productivity uplifts of 30 to 40 percent on comparable workloads.

On professional services, a Harvard Business School study published in late 2023 examined 758 management consultants at a single global strategy firm. The consultants using AI completed 12 percent more tasks on average, completed those tasks 25 percent faster, and produced work that external judges rated 40 percent higher in quality. The gap between the most capable and the least capable consultants narrowed considerably in the AI-augmented condition — another data point confirming that AI lifts the bottom of the performance distribution faster than it lifts the top.

On finance and accounting, which is the function I watch most closely given my firm’s core practice, the evidence is more fragmented but pointing in the same direction. Deloitte’s 2025 global CFO signals survey reported that finance functions with mature AI deployment were closing their monthly books approximately 30 percent faster than comparable peers without such deployment. KPMG’s 2025 global AI in finance study, drawing on interviews with more than 400 organisations, reported average reconciliation processing time reductions of 50 percent in AI-enabled environments. Independent controls testing — the specific work my firm does for audit clients — is being accelerated by agentic tooling in ways that were not available even eighteen months ago.

I could go on, and in subsequent articles of this series I will. The point of listing these studies is not the individual numbers. The point is the consistency of the direction and the size of the effect. When productivity studies, conducted by independent researchers, covering independent function areas, converge on effects in the 20 to 50 percent range, that is not an artefact of measurement methodology. That is a real signal about what this technology is doing to the economics of professional work.

 

WHAT THIS MEANS IN CARIBBEAN TERMS A 30 percent productivity gain across the professional-services layer of a Caribbean economy is not a cost-reduction story. It is a competitiveness story. A Jamaican credit union that captures even half of that gain will be able to offer equivalent member service at lower cost, or superior member service at equivalent cost — and its members will notice. A Trinidadian manufacturer that captures the full gain will have a different cost structure than its regional competitors within two years. The effect compounds because the firms that move early build the governance, the workforce familiarity, and the data estate that make subsequent AI investments cheaper and lower-risk for them. The firms that move late face the same costs at a later date and without any of those compounding advantages.

Why the Caribbean lag pattern will not hold this time

Caribbean enterprises have historically lagged global technology adoption cycles by between eighteen and thirty-six months. This is not a criticism of our region’s executives; it is a factual observation about market structure. Our markets are smaller, our capital is more expensive, our talent pools are narrower, our regulatory frameworks are more fragmented, and our technology vendors are further away. A chief executive in Kingston or Bridgetown or Port of Spain who waits for a new technology to mature before committing to it is usually behaving rationally, not slowly.

In most technology cycles, this lag is manageable. The firms that move late pay a premium in missed opportunity, but they avoid the costs of bleeding-edge deployment. The catch-up period is typically long enough to be absorbed into the normal investment rhythm of a well-run business.

The agentic cycle will not be manageable in this way, and I want to specify exactly why.

First — the compounding is faster than in previous cycles

An organisation that deploys agentic tooling meaningfully in 2026 spends that year training its workforce, building its governance discipline, and populating its data estate with the artefacts that AI systems actually need to do useful work. By 2027, that organisation is operating with a meaningfully different cost base than its peers. By 2028, the gap between the adopters and the non-adopters is visible in market share, in customer retention metrics, and in the calibre of talent being recruited into the two groups. The catch-up cost in 2028 is not the 2026 cost with two years of inflation added. It is the 2026 cost plus the compounded advantage the adopters have built — which is a much larger number than simple inflation would suggest.

Second — the talent becomes scarce

The Caribbean workforce capable of governing, deploying, and sustaining AI in a regulated enterprise environment is small today. It will be smaller in 2028, not because the supply will have shrunk but because the demand will have exploded. Firms that build this capability now will have it. Firms that try to buy it in 2028 will find that the people capable of the work have already been contracted to the firms that moved earlier. This is not a theoretical concern. It is the pattern I watched play out with cloud architects in our region between 2015 and 2020, and it is the pattern I watched play out with cybersecurity professionals between 2019 and 2023. In both cases, the firms that built internal capability during the build-up phase had a material cost advantage for years afterwards. The firms that tried to buy capability during the peak phase paid two or three times the market rate, and many still could not find the talent they needed.

Third — the data gap widens

AI agents work against an organisation’s data, its documents, and its institutional knowledge. The quality of work these systems can do is bounded by the quality of the information they can reach and trust. Organisations that begin the work of inventorying, classifying, and structuring their data estate in 2026 are doing something useful in its own right — and they are also, as a second-order effect, making their organisation more amenable to every subsequent AI deployment. Organisations that do not begin this work will discover in 2028 that they are not simply behind on AI. They are behind on the foundations that would make AI useful to them. Fixing the foundations takes years, not quarters.

Fourth — the regulatory surface is hardening

Caribbean regulators have been watching the European Artificial Intelligence Act, the emerging United States federal frameworks, and the early regional guidance issued by regional central banks and data protection commissioners. The consensus in the regulatory conversations I participate in privately is that meaningful AI-specific regulation will land in most major Caribbean jurisdictions within the next thirty-six months. Firms that build governance maturity ahead of the rule will shape the rule. Firms that wait for the rule will inherit it. Inheriting a regulatory framework is always more expensive than helping to shape one, and in the case of AI regulation the gap between those two positions is likely to be particularly wide, because the cost of retrofitting governance onto already-deployed AI systems is substantial.

 

Firms that shape the regulatory framework will prosper under it. Firms that wait for it will inherit it — and inherited frameworks are always more expensive than shaped ones.

These four factors do not act independently of each other. They compound. An organisation that moves in 2026 is simultaneously building governance maturity, accumulating deployment experience, training its workforce, and structuring its data estate — and each of these activities feeds the others. An organisation that tries to move in 2028 faces four simultaneous uphill climbs, in a market where the best resources have already been committed elsewhere, in a regulatory environment it had no voice in shaping.

This is the specific reason I say the normal Caribbean lag pattern will not hold in this cycle. The window for timely action is genuinely narrower than in previous technology cycles, and the cost of missing it is genuinely steeper.

The cloud precedent — what the last delay actually cost

I made a claim at the start of this article that the Caribbean is still paying for the cloud migration it failed to complete in a timely way between 2013 and 2018. Because the argument that follows depends on that claim being credible, let me substantiate it.

Between 2013 and 2018, enterprise cloud adoption in North America and Western Europe moved from the early-majority phase into the late-majority phase of the classic technology adoption curve. Average cloud spend as a proportion of enterprise IT spend crossed the 20 percent threshold in those markets by 2016. The comparable number in the English-speaking Caribbean did not cross 20 percent until approximately 2020, and only on the most generous interpretation of the available data. Our region trailed the adoption curve by approximately four years.

The consequences of that four-year trail are not theoretical. Caribbean enterprises today carry, on average, more legacy infrastructure cost than their North American peers; spend a higher proportion of their IT budgets on maintenance of existing systems rather than on new capability; have smaller internal benches of modern engineering talent; and are more exposed to the cybersecurity vulnerabilities that come with on-premises technology estates that were hardened for a threat environment that no longer exists. These are not catastrophic positions. But they are real competitive disadvantages that compound quarter after quarter. The Caribbean firms that closed the gap between 2018 and 2022 are in materially better operational shape today than the Caribbean firms that are still closing it in 2026.

 

A PUBLICLY OBSERVABLE PATTERN The Jamaica Stock Exchange currently lists approximately forty-six companies on its main market. Of those, fewer than ten have made any public reference to artificial intelligence in their 2024 annual reports, and no listed Jamaican company has yet published a formal AI governance framework as part of its corporate-governance disclosures. By comparison, approximately 70 percent of the S&P 500 constituents disclosed material AI activity in the same reporting cycle, and roughly one in three published, in some form, a governance position on the technology. This is not a criticism of Caribbean disclosure practice. It is a measurement of where our region’s public markets currently sit on the AI maturity curve, and a useful baseline against which we can assess, in subsequent years, whether the region is catching up or falling further behind.

The cloud delay was survivable because cloud itself was a fifteen-year story, and the catch-up window was correspondingly long. The agentic delay will not be a fifteen-year story. The agentic delay is playing out in real time, with enterprise deployment moving from experiment to production between 2024 and 2026. The catch-up window is measured in quarters, not in years. Organisations that begin the work this year will have meaningful maturity by the end of 2027. Organisations that begin in 2028 will be working to catch up with a moving target, against competitors who have already built the foundations, in a talent market that has tightened, inside a regulatory framework they had no voice in shaping.

I do not want to be melodramatic about this. The Caribbean economy has survived every previous technology delay we have experienced. The region is resilient, its executives are capable, and in some sectors — parts of our financial services industry in particular — we have leapfrogged from earlier generations of technology directly to current ones. None of what I have written above is an argument that the Caribbean is doomed. It is an argument that the cost of delay is higher in this cycle than it has been before, and that the margin for the normal Caribbean lag pattern is narrower. That is all it is, but that is enough.

A diagnostic for your leadership team this quarter

I want to close this article with something more practical than an argument. I want to give you a specific instrument you can use in the next ninety days to take the measure of your own organisation’s AI position — an instrument that requires no external consultant, no formal assessment, and no technology evaluation. It is a set of three questions, asked in sequence, to the right people, in the right setting. I call it the D-AGENTICA™ Three-Question Board Diagnostic. It is the first in a series of named diagnostic instruments that will appear in this series.

The structure of the instrument is simple. The three questions are deliberately obvious in their surface form. What matters is not the question itself; it is the pattern of answers. The pattern reveals, more precisely than any maturity assessment tool I have seen, where a specific Caribbean organisation actually sits on the AI readiness curve.

Ask each question in an individual conversation rather than a group meeting. Group dynamics suppress the honesty that this instrument depends on. Ask them in sequence, over a period of two to three weeks. Write down the answers. Then look for the patterns described below.

A NAMED INSTRUMENT The D-AGENTICA™ Three-Question Board Diagnostic Three questions a Caribbean CEO can ask their leadership team this quarter that will reveal more about the organisation’s AI position than any external consultant could in six weeks. Use them in sequence, in individual conversations before group ones, and listen carefully to the structure of the answers. QUESTION 1 Where are we on AI today, and who owns it? What it reveals: Accountability gaps. If three different people believe they own parts of it and no one believes they own all of it, you have discovered your highest-priority organisational gap. QUESTION 2 What AI tools are we using that we have not sanctioned? What it reveals: Shadow-AI exposure. Most Caribbean organisations have meaningful shadow adoption of consumer AI tools. A written inventory from the head of technology within ten working days will tell you how much of that is governed. QUESTION 3 What could we do better with AI in the next twelve months that we are not doing? What it reveals: Leadership AI literacy. Ask each executive individually. Specific, numerical answers indicate genuine engagement; generic answers about ‘innovation’ and ‘disruption’ indicate the opposite. The pattern of answers is your leadership readiness map.

Once you have asked all three questions of all of your senior executives, you will have surfaced three things. First, you will know whether your organisation has meaningful, accountable AI ownership — or whether ownership is spread in a way that produces the appearance of coverage without the substance of accountability. Second, you will know the scale of the shadow-AI adoption already running inside your organisation, which is almost certainly larger than leadership currently believes. Third, you will know which of your senior executives are thinking seriously about AI and which are performing the appearance of engagement.

In my experience advising Caribbean institutions across the last three years, executives who complete this diagnostic honestly discover that their organisation is further behind than they had believed — and that most of the gap is a leadership gap rather than a technology gap. I consider this good news. Technology gaps are expensive to close. Leadership gaps are inexpensive to close, once you have noticed them, because the only thing required to close a leadership gap is a decision.

What we have established, and what comes next

In this first article of the series, I have tried to do three things. I have argued that the current technology cycle is genuinely different from those that came before it, because the systems arriving now are not tools but agents — software that plans, acts, and completes work against goals. I have shown that independent evidence converges on productivity effects in the 20 to 50 percent range across the major functions of professional work. And I have set out the four specific reasons why the normal Caribbean lag pattern will not provide cover in this cycle: the compounding is faster, the talent is scarcening, the data gap widens, and the regulatory surface is hardening.

I have not argued that the Caribbean is doomed. I have argued that the cost of delay is higher in this cycle than in previous ones, and that the margin for our usual lag pattern is narrower. I have given you an instrument you can apply to your own organisation this quarter, and I will expand on that instrument and introduce others as this series progresses.

The next article, will do something specific and necessary. It will demystify what AI agents actually do inside an enterprise. Most Caribbean executives have used consumer AI tools — they have typed a question into ChatGPT or Claude and received an answer. Very few have a clear mental model of the difference between a chatbot, a copilot, and an agent, and that gap is the single largest source of confusion in current Caribbean boardroom conversations about AI.

The next article installs the vocabulary. It goes through what an agent actually does, what four things it requires, what it does not do, and — critically — how to tell a genuinely agentic product from a chatbot that has been rebranded as an agent for marketing purposes. We will work through four concrete examples of AI agents in Caribbean contexts, including an agent for audit workpaper preparation, an agent for GCT compliance workflows, an agent for credit union member service, and an agent that functions as a Virtual CFO. By the end of that article, you will be able to walk into any vendor meeting and assess, within five minutes, whether the product being pitched to you is actually agentic.

If you have found the argument in this article persuasive, I want to ask one specific thing of you. Share this article with one other person on your board or executive team. The single most consistent pattern I have observed in the Caribbean organisations that are moving well on artificial intelligence is that two or three senior people reached independent conviction at roughly the same time and were able to carry the rest of the organisation with them. The single most consistent pattern I observed in the organisations that fell behind in the cloud migration was that one senior person saw the point early and could not persuade the others until it was too late. The cost of persuading one more colleague now is a five-minute conversation. The cost of not persuading them is measured on a different scale entirely.

 

FOR THE BOARD AGENDA This article has made a specific argument about why the current AI cycle matters for Caribbean enterprises and why the time available for timely action is narrower than usual. A board chair or audit committee chair reading this article has earned the right to ask their leadership team two specific things. THE QUESTION Have we genuinely engaged with the question of whether artificial intelligence will change the competitive position of our organisation in the next thirty-six months — and if we have, what is our actual position, expressed in a form our board can test? THE DECISION That the executive team will apply the D-AGENTICA™ Three-Question Board Diagnostic across our senior leadership within the next sixty days, and will report the findings to the board at its next regularly-scheduled meeting, with a proposed initial position on artificial intelligence for board consideration.

 

THE CARIBBEAN AI ADOPTION IMPERATIVE A 12-Article Series from Dawgen Global NEXT IN THIS SERIES Article 02 — Beyond the Hype What AI agents actually do in an enterprise MEASURE YOUR ORGANISATION’S AI READINESS Request the free D-AGENTICA™ AI Maturity Self-Assessment Email : [email protected]

About Dawgen Global