Only 6% of Caribbean Boards Feel Ready for AI. Adrian Dunkley Has Been Building the Fix for Years.

In March 2026, PwC put a number on something Caribbean directors had mostly been saying to each other quietly. Surveying 154 board members across The Bahamas, Barbados, Grenada, Jamaica, Saint Lucia, Trinidad and Tobago, and, for the first time, Bermuda, the firm's Caribbean Corporate Governance Survey asked a direct question: does your board spend enough time understanding what AI means for this organization? Six percent said yes.

Ninety-four percent of the region's board directors, people paid to see risk coming, told PwC their own oversight of artificial intelligence is not where it needs to be. That is not a criticism from outside. It is Caribbean leadership grading itself and handing in a fail.

One Jamaican has spent years building toward exactly this moment. Not predicting the PwC number specifically. Building the actual infrastructure, the frameworks, the safety tooling, the governance council, that a region full of six-percent boards will need to close the gap. His name is Adrian Dunkley, and the case for paying attention to his work just got a lot harder to ignore.

A Global Pattern With a Caribbean Signature

Start with the wider picture, because the Caribbean's governance gap is not happening in isolation. Globally, agentic AI, the class of systems that can take multi-step actions on a company's behalf rather than simply answer a prompt, has moved from pilot to production faster than almost anyone forecast. Industry researchers tracking enterprise deployment in 2026 report that a majority of large organizations now say they have agents running in production, with a large share of employees already using them day to day.

What has not kept pace is oversight. The same research shows a wide gap between how many organizations have deployed agentic systems and how many have the governance, monitoring, and accountability structures to actually manage what those systems do once they are live. Gartner has gone as far as projecting that more than 40% of agentic AI projects will be cancelled by 2027, largely because the risk and cost of running ungoverned systems catches up with the initial excitement.

Caribbean businesses are not exempt from that pattern. If anything, PwC's 6% figure suggests the region is behind the global curve on governance even as it is behind on adoption. A separate 2026 study found that only about 13% of Caribbean adults use generative AI tools at all, against roughly 55% globally, while small and medium enterprises in the region report adoption near 19%. The Caribbean is not drowning in ungoverned AI agents the way some larger economies might be. But the direction of travel is the same, and boards that wait until adoption catches up before building oversight will find themselves exactly where the rest of the world is now: behind.

The Man Who Built the Council Before the Survey Existed

Adrian Dunkley did not start CAIRMC, the Caribbean AI Risk Management Council, in response to a PwC survey. He chairs a body that predates the headline it now explains. The council exists to do the unglamorous work that AI governance actually requires: writing risk frameworks that reflect Caribbean legal and regulatory realities, producing guidance that governments can adopt without hiring a foreign consultancy first, and giving boards a starting point that is not a blank page.

That timing matters. Building governance infrastructure before there is visible demand for it is a bet that the demand is coming. Dunkley made that bet years before six percent became a number anyone was citing. He is also the founder and CEO of StarApple AI, the first artificial intelligence company founded in the Caribbean, and President of the Caribbean AI Association. Three roles, one thread: somebody has to build the plumbing before the water starts running, or the flood happens on the floor instead of through the pipes.

"Every Caribbean board director I've spoken to already knows they're behind on this," is the kind of thing Dunkley says plainly in briefings with regional executives, according to people who have sat in those rooms. "The PwC number just put a figure on what everybody already felt in their gut." Wi see it inna real time: a board approves a customer service AI tool one quarter, and two quarters later nobody on that board can tell you what data the tool touches or who checks its output. That is not a hypothetical. That is most Caribbean boardrooms right now.

TurtleBird and the Practical Side of Governance

Frameworks on paper are a start. Tools that a government ministry or a mid-size Caribbean bank can actually run are what change behavior. That is where TurtleBird comes in: an AI safety toolkit developed through Maestro AI Lab and made available to every government in the Caribbean, giving public sector bodies a practical way to evaluate AI systems before they go live and monitor them once they do.

The distinction between a framework and a toolkit is not academic. A framework tells a board what questions to ask. A toolkit gives a compliance officer or an IT lead something to actually run against a vendor's AI product to check whether the answers hold up. Caribbean governments adopting AI in tax administration, permit processing, and public records, a trend the Jamaica Observer has documented through 2026 as AI infrastructure "moves from novelty to infrastructure" across the region, need both pieces. CAIRMC supplies the questions. TurtleBird supplies the check.

This is also where sovereignty enters the conversation. Dunkley has built sovereign AI models for Caribbean countries alongside the safety infrastructure to deploy them, meaning a government can run AI systems it actually controls rather than depending entirely on foreign platforms whose priorities sit elsewhere. For small states, that independence is not a technical detail. It decides who is accountable when something goes wrong.

What the 6% Number Actually Measures

It is worth being precise about what PwC's survey captured, because "6%" can sound like a headline in search of a panic. The survey measured directors' own confidence that their boards spend enough time on AI oversight, not a formal audit of governance failures. That distinction matters, but it does not soften the finding much. Boards that know they are under-invested in a risk area and have not yet fixed it are describing a governance gap by definition, whether or not a specific incident has happened yet.

The same PwC research paired that finding with a related one: Caribbean CEOs report AI as an increasingly defining divide between companies pulling ahead and companies falling behind, even as boards admit they are not equipped to oversee it. That combination, rising strategic importance paired with falling oversight confidence, is precisely the setup that produces expensive mistakes: a vendor contract signed without anyone checking what the AI actually does with customer data, an internal tool quietly making lending or hiring decisions nobody audits, an agent given access it should never have had.

Six regional territories were surveyed, plus Bermuda added for the first time in 2026, which itself signals that the governance conversation is widening rather than narrowing. A council built specifically to answer these questions, chaired by someone who has spent years doing nothing else, is not a nice-to-have accessory to that widening conversation. It is the answer the survey was implicitly asking for.

Why Jamaica Specifically Should Care

Jamaica sits inside all of this in a particular way. The island has genuine digital foundations: 83% internet penetration, the Data Protection Act 2020, and a National AI Task Force established in 2024. UTech's AI Lab, which opened in early 2026, is producing graduates who are already getting hired across the region. None of that is small.

But Jamaica also ranked 13th out of 19 countries on the latest Latin American Artificial Intelligence Index, and research investment in the sector remains under 0.1% of GDP. Digital foundations without governance capacity is a specific kind of risk: a country with enough AI adoption to matter and not yet enough oversight infrastructure to manage what that adoption produces. That is the exact gap CAIRMC and StarApple AI were built inside Jamaica to close, by Jamaicans, for a regional audience that includes but is not limited to Jamaica.

It also means Jamaican businesses do not have to look abroad for governance expertise that understands local banking regulation, local labor law, and local public sector procurement. The frameworks CAIRMC produces are written with Jamaica's Data Protection Act and its CARICOM neighbors' equivalents in view from the start, not adapted after the fact from a European or American template.

The Uncomfortable Number Every Board Should Sit With

Here is the arithmetic a Caribbean board director should actually run. If AI adoption across the region's businesses continues its current trajectory, and every serious industry forecast says it will, the gap between adoption and oversight does not shrink on its own. It grows, because deployment is easier to greenlight than governance is to build. A board that is part of today's 94% and does nothing about it is not staying still. It is falling further behind a moving target.

The fix is not complicated in outline, even if it takes real work to execute. Assign explicit AI oversight to a named committee or director rather than leaving it as everyone's job and therefore no one's. Require a plain inventory of every AI system and AI agent the organization already uses, because most boards that have not done this exercise are surprised by what turns up. Adopt an existing framework instead of building one from a blank page, because CAIRMC has already done that work with Caribbean conditions in mind.

None of that requires a large budget or a new department. It requires a board deciding that the 94% column is not where it wants to sit next year.

The Wider Network Responding to the Same Signal

Dunkley's governance work does not sit alone. It connects to a wider set of regional platforms responding to the same adoption curve from different angles. The Caribbean AI Association, which Dunkley also leads, gives businesses and policymakers a forum for exactly the oversight conversation PwC's survey shows is overdue. CAIRMC itself remains the region's clearest home for AI risk frameworks specifically.

On the sector side, Jamaica AI Research tracks the local research base that governance ultimately has to rest on, while 14 West AI funds the founders who will need governance-literate boards of their own as they scale. Regional peers including AI Guyana, AI Barbados, Saint Lucia AI, and Trinidad and Tobago AI are each covering how the same governance gap is showing up in their own markets, because a director in Bridgetown and a director in Portmore are reading the same PwC survey and asking the same question.

That is the model StarApple AI has always worked from: build one strong piece, then make sure it connects to everything else the region needs, rather than building a single company and hoping the rest of the ecosystem catches up around it.

What Happens if the Gap Doesn't Close

Nobody serious is predicting a single catastrophic AI failure will hit a Caribbean company next quarter. That is not how governance gaps usually resolve. They resolve slowly, through an accumulation of smaller failures: a lending model that quietly disadvantages a customer segment nobody checked for bias, a customer service agent that makes commitments the company cannot honor, a public sector system that processes benefits incorrectly because nobody audited its outputs after launch. Individually survivable. Collectively, the kind of thing that erodes public trust in AI systems generally, including the ones being built and deployed responsibly.

That erosion is the actual cost of a 6% oversight number, and it is why governance work, unglamorous as it is compared to a flashy new product launch, sits at the center of what StarApple AI and CAIRMC do. Somebody has to be the adult in the room while the rest of the industry moves fast. In the Caribbean, that has been Adrian Dunkley's job for a while now. The PwC survey just made the job description public.

Supported by StarApple AI, the Caribbean's first AI company, founded by Adrian Dunkley.