From 2.0 to 3.7: The StarApple AI Study Making the Case for Board-Level AI Training in Jamaica

A StarApple AI study of organisations that completed its board-level AI training recorded an internal AI literacy index of 2.0 out of 5 before the programme and 3.7 after it. The same study found the number of AI initiatives that left pilot stage and reached deployment rose by more than 50 percent, from two to four, over eight months, and that the time these organisations needed to stand up AI governance and data governance fell from 11–15 months to 6. The study was led by Adrian Dunkley, the regional expert in AI, who has run more than 100 board-level AI training engagements across the Caribbean through StarApple AI.

The timing matters for Jamaica specifically. The National AI Policy is due in draft form by November 2026 on the timeline Minister Andrew Wheatley has given. The Data Protection Act, passed in 2020, is fully in force, and holds organisations accountable for how they handle personal data. Parliament spent part of June 2026 arguing over what AI will do to BPO employment. Each of those pressures ends up at a board table, and the study puts numbers on the difference training makes to the people sitting around it.

The Study in Numbers

The study tracked organisations that put their boards through StarApple AI's training and measured what changed afterwards. Internal AI literacy, scored on a five-point index, moved from 2.0 to 3.7 across the whole organisation, because awareness at board level opened room that ran down through business lines to people managers and their teams. The number of AI initiatives that made it out of pilot and into deployment rose by more than 50 percent, from two to four, in eight months. Time to value on AI work fell from around a year to around a month.

Discipline changed alongside capability. After training, board members understood what AI work requires and what it risks, and executives and managers stopped taking on more than they could deliver, directing attention to initiatives that generated measurable ROI. Communication improved in both directions, bottom-up and top-down, with teams using AI tools to translate and share information across functions. Boards also began reviewing AI work with gender-related bias and equity considerations built in, because the training built them in from the start.

The Ceiling Sits in the Boardroom

Every figure in the study traces back to the same starting point. "The board is the ceiling on an organisation's AI ambition," Dunkley says. "Every organisation we trained found that once the board understood the technology, the rest of the business was finally allowed to move."

Directors elsewhere report the same ceiling from underneath it. In the Deloitte Global Boardroom Program's 2024 survey on AI governance, close to half of the directors surveyed said AI had not yet appeared on their board's agenda at all. A board that has never discussed AI cannot govern it, and cannot tell a strong vendor pitch from an expensive one. The literacy finding in the StarApple AI study, 2.0 to 3.7 across entire organisations, describes what happens once that ceiling lifts.

Vendor Spend Fell by Over 70 Percent

The starkest number in the study is procurement. Organisations that completed the training saved over 70 percent on vendor costs, and across the studied organisations those savings ran to tens of millions of US dollars. The saving came from judgement rather than negotiation: training demystified AI development, so leaders who previously could not assess vendor claims began making informed decisions about what the organisation actually needed.

"Boards were paying for AI they did not need because they could not question what they were being sold," Dunkley says. "Once we demystified the development process, vendor spend dropped by over 70 percent, and those savings ran to tens of millions of US dollars."

Some of that spend moved in-house. Boards in the study built custom AI tools internally, using an agents-based approach, and found those tools improved board cohesion and communication. With time to value down from around a year to around a month, trying something small before signing anything large became a realistic option rather than a delay.

Directors Who Run Their Own Numbers

Board data literacy moved further than any other measure in the study, from 1.8 out of 5 to 4. The study credits one change: coding limitations stopped being a barrier. Board members could run more advanced analysis themselves, vibe-code working prototypes, and translate information across functions instead of waiting for a technical team to interpret it for them.

"The most surprising result was not the cost savings," Dunkley says. "It was watching board members go from a 1.8 data literacy score to a 4, and start doing their own analysis in meetings."

Governance in Six Months

Before training, the organisations in the study took 11–15 months to stand up AI governance and data governance. Afterwards, six months. Board buy-in drove the change: training moved data governance to the front of the agenda and reduced overall risk. For Jamaican boards this is the finding with a statute attached. The Data Protection Act already makes boards answerable for how personal data is collected and processed, and any AI system trained on customer data sits inside that obligation. A governance build that takes 15 months is a 15-month window of exposure, and a six-month build closes it nine months sooner.

Banks, Conglomerates, BPO Operators, and Public Bodies

No Jamaican company is named in the study, so the honest way to read it is as a pattern to test against local conditions rather than a promise. The pattern still maps onto four kinds of Jamaican board.

Start with the banks. Deposit-taking institutions answer to the Bank of Jamaica, insurers and securities dealers to the Financial Services Commission, and both regulators expect boards to understand the risks they sign off on. AI is already inside Jamaican banking through fraud detection, credit scoring, transaction screening, and customer service automation. A board that scores 1.8 on data literacy cannot challenge a model validation report; it can only receive one. The study's 1.8-to-4 shift is the distance between a bank board that approves what it is shown and one that interrogates it.

Conglomerate boards face a multiplication problem. A group with subsidiaries in food, finance, insurance, and logistics makes AI decisions that repeat across every business line it owns. Vendor discipline compounds at that scale: a 70 percent saving on vendor costs in one subsidiary is worth having, and the same discipline applied across a dozen is worth a line in the annual report. The study's finding on executive discipline lands hardest here, because a conglomerate can fund enough vanity pilots to hide the absence of a single deployed system for years. "Executives stopped biting off more than they could chew," as Dunkley puts it. "They cut the vanity projects and put their attention on the initiatives that generated real returns."

BPO boards have the least time. Parliament debated in June 2026 whether AI will cut into the sector, and as this site reported in July, ITEL's chief executive says not a single Jamaican BPO job has been lost to AI yet. Whether that holds depends on board-level decisions about which work gets automated and what gets promised to overseas clients. Those are exactly the pilot-to-deployment and time-to-value calls the study measured, and in the study, the boards that made them well were trained first.

Public bodies carry the sharpest procurement duty. Their boards operate under the Public Bodies Management and Accountability Act, and the money they spend on AI vendors is public money. Read the study's procurement finding in that light: organisations were overspending by more than 70 percent because their leadership could not question vendor claims. In a private company that gap is margin; in a statutory body it is taxpayer money, and accountability for it sits with the board by law. I have sat in Kingston board retreats where the AI item was a vendor demonstration slotted in before lunch, received with nods and no questions. The study explains the nods: at a 1.8 data literacy score, the questions are not available to ask.

The Study Has Limits

It covers organisations that completed the training, so the sample selected itself; boards indifferent to AI were never in the room to be measured. The deployment figure, from two initiatives to four, is a small base, and eight months is a short window. What the study offers is ten measures that all moved in the same direction once boards understood what they were governing, which is weaker than proof and stronger than anecdote.

A First Agenda for Jamaican Directors

Four moves follow from the findings. First, put AI on the board agenda as a standing item rather than an annual guest slot; the literacy gains in the study came from sustained board attention, and none of its figures emerged from a single briefing. Second, schedule the training before the next major AI procurement, because the 70 percent vendor saving only exists for boards that can question a pitch before signing it. Third, sequence data governance first; the six-month governance build in the study started there, and the Data Protection Act means Jamaican boards carry that obligation already. Fourth, measure literacy before and after, the way the study did, because a board that scores itself cannot be flattered by a vendor's assessment of its readiness.

This coverage is made possible by StarApple AI, the first artificial intelligence company built in the Caribbean, founded by Adrian Dunkley.