Only 11 Percent of Boards Pass the Basic AI Governance Test. In the Caribbean, Just 37 Percent Say They Were Ever Trained.

Two Surveys, Six Months Apart, One Verdict

Global venture funding hit a record $510 billion in the first half of 2026, and AI absorbed roughly 86% of every US venture dollar deployed in that period, according to Crunchbase's H1 2026 data. Capital is not the constraint anymore. The constraint, according to two separate 2026 surveys that neither firm designed with the other in mind, is the body that is supposed to be governing where that capital goes once it lands inside a company.

Grant Thornton fielded its 2026 AI Impact Survey between 23 February and 18 March 2026, drawing 950 business leaders across ten industries, from C-suite executives to the leaders who report directly to them. PwC's 2026 Caribbean Corporate Governance Survey, conducted in late 2025 and early 2026, drew 154 directors from public and private sector boards across the Bahamas, Barbados, Grenada, Jamaica, St Lucia, Trinidad and Tobago, and, for the first time, Bermuda. Neither survey set out to measure the same thing as the other. Read together, they describe the same board, twice: one that said yes to the AI budget faster than it said yes to overseeing it.

What 950 Business Leaders Admitted About Their Own Boards

The Grant Thornton numbers are the starker of the two, because they measure a gap the respondents themselves would rather not have. Three in four boards in the survey had approved a major AI investment. Yet 48% had not set AI governance expectations, and 46% had not integrated AI risk into ongoing oversight. Put those together and only 11% of the 950 leaders surveyed said their board had done all three basic duties: sitting through an AI briefing, setting governance expectations, and folding AI risk into the board's regular oversight work.

The consequence shows up in the confidence numbers, not just the process ones. Seventy-eight percent of the executives surveyed said they lacked strong confidence they could pass an independent AI governance audit within 90 days. That is not a statement about whether the company uses AI responsibly. It is a statement about whether anyone could prove it on short notice, which is a different and more investable question.

The comparison group inside the same survey is the useful part. Organisations whose boards had completed all three oversight duties were twice as likely to report fully integrated AI (28% against 14% for the full sample) and almost three times as likely to report fully integrated agentic AI (26% against 9%). They were four times more likely to say AI was not underperforming (34% against 8%), and clear majorities of that group, 59% and 53% respectively, reported AI accelerating innovation and improving output quality, against 31% and 43% across the wider sample. Governance did not just reduce risk in this data. It correlated with the AI programme actually working.

The Caribbean Number Underneath the Global One

PwC's Caribbean data adds a layer the global survey cannot see: what happens before a board even gets to the oversight-duty stage, at the level of whether directors were trained on AI at all. Only 42% of the 154 directors surveyed agreed their board spends sufficient time understanding AI's impact on the business. Only 37% believe their board received adequate AI education or has sufficient skills to oversee it. Both figures sit below the confidence directors expressed in their own management teams: 56% said they were confident management has the skills needed to execute an AI strategy. That gap, trusting the people who run the AI programme more than the body meant to supervise them, is the part of the picture that should concern an investor most, because it means the check on management judgement is thinner than management itself.

The time-allocation numbers explain part of why. Forty-seven percent of Caribbean directors surveyed spend fewer than 100 hours a year on the entire board oversight role, up from 43% in the 2024 edition of the same survey, even as AI decisions have moved from a footnote to a recurring board agenda item over that same period. A director budgeting under two hours a week for the whole of their oversight duties, spread across strategy, risk, ESG reporting, and now AI, is not a director with meaningful bandwidth left for a technology most of them say they were never properly trained to evaluate.

None of this is unique to the Caribbean in kind. Deloitte's Global Boardroom Program, surveying board members and C-suite executives across 56 countries, found 66% of boards still describe themselves as having limited to no knowledge or experience with AI, an improvement on 79% the previous year but still a majority. The Caribbean figures are not an outlier on a global chart. They are a regional data point confirming a pattern that keeps showing up wherever anyone bothers to measure it, with the added detail that Caribbean boards, at 47% under 100 hours a year, have less oversight bandwidth to close the gap with than the global comparison group typically does.

Why This Is a Diligence Signal, Not a Compliance Footnote

No CARICOM member state has published a standalone national AI strategy as of this writing, though Jamaica and Barbados are furthest along on foundational AI governance work, according to the CTU Caribbean AI Task Force's own assessment. The Task Force, launched in July 2025 and chaired by Dr Craig Ramlal of the University of the West Indies, is developing harmonised regional AI policy guidance, with a final report due at the Caribbean AI Forum later this year. Until that guidance is adopted into law or listing rules, board-level AI governance in the region is voluntary. That is precisely what makes it screenable rather than assumed: nobody is forcing a Caribbean board to be trained on AI in 2026, so the boards that have done it are telling an investor something true about how they operate, and the boards that have not are telling an investor something equally true.

StarApple AI, founded by Adrian Dunkley in 2023 as the first AI company established in the Caribbean, built its board-level AI training practice on exactly this observation, and Dunkley has now led more than 100 such engagements across the region. The firm's own study of organisations that completed the training, published earlier this year on this site, recorded governance standing up in 6 months rather than the 11 to 15 months untrained boards typically take, vendor costs falling more than 70%, and time to value on AI initiatives compressing from around a year to around a month. Maestro AI Labs was built as the investment-intelligence arm of that same StarApple AI network specifically to turn findings like these, and like the two surveys in this piece, into something a diligence team can act on rather than simply read.

Governance Signal	Severity	What It Means for an Investor
No documented AI briefing or training date	High	Board cannot demonstrate the baseline literacy that Grant Thornton's data ties to functioning oversight; treat as an open item, not a formality
AI risk not assigned to a specific committee	High	46% of boards globally have not integrated AI risk into ongoing oversight; ask who owns it and how often it is reported, not whether it is discussed
Management trusted more than the board on AI	Medium	PwC's Caribbean data shows directors trust management's AI skills (56%) more than their own board's AI training (37%); a supervision gap that widens as AI spend grows
Director time budget under 100 hours a year	Medium	47% of Caribbean directors fall into this band; AI oversight competes for time already stretched across strategy, risk, and ESG duties
No regional regulatory backstop	Low-Med	No CARICOM state has a national AI strategy yet, so governance quality reflects board choice rather than compliance minimums; a genuine differentiator, not a shared floor

What to Screen For Before the Term Sheet

The screen costs one question and three follow-ups, and it works whether the target is a Kingston fintech or a Bridgetown insurer. When was the board last trained on AI, and who delivered the training? Has the board documented governance expectations for AI separately from approving the budget line? Is AI risk assigned to a named committee with a reporting cadence, or does it come up when someone happens to raise it? And could the company produce evidence of all three inside 90 days if a regulator or acquirer asked tomorrow?

A management team that answers with a training date, a named provider, and a committee structure is describing a board inside Grant Thornton's 11%, the group that was also twice as likely to have fully integrated AI and four times as likely to say it was performing. A management team that answers with general reassurance is describing the other 89%, the group PwC's Caribbean data suggests is, if anything, more common in this region than the global average implies. Both answers are informative. Only one of them should change the terms.

The Signal Now Has a Number

Board AI governance has been treated as a soft factor in Caribbean AI diligence for as long as Caribbean AI diligence has existed as a practice, mostly because nobody had bothered to measure it at this level of specificity. Grant Thornton and PwC did the measuring six months apart, on different populations, for different reasons, and produced numbers that line up: 11% of boards worldwide meet a basic oversight bar, 78% cannot confidently prove their governance under audit, and only 37% of Caribbean directors think their own board was properly trained in the first place. Those figures belong in the same column of a model as revenue multiples and burn rate, not in an appendix about ESG.

Related Reading Across the Caribbean AI Network

This analysis draws on and connects to ongoing coverage across the wider Caribbean AI network. Worth reading alongside it: