Board Data Literacy Went From 1.8 To 4 Out Of 5: Inside StarApple AI's Board-Level AI Training Study

A StarApple AI study of organisations that completed its board-level AI training measured directors' data literacy before and after the programme. The average score moved from 1.8 out of 5 to 4 out of 5. A board at 1.8 reads its numbers second-hand, through whoever prepared the pack. A board at 4 opens the data itself. In the organisations the study followed, directors ran their own analysis during meetings and built working prototypes of tools they had previously only been able to describe.

Adrian Dunkley, the Caribbean's leading AI expert, led the study, and has run more than 100 board-level AI training engagements across the region through StarApple AI. StarApple Analytics is a subsidiary of StarApple AI, and we build the dashboards, models and pipelines that boards like these end up reviewing, so the finding lands close to home. It also answers a question our own industry has been nervous about since generative coding tools arrived: once the people at the top can do basic analysis themselves, demand for professional analytics grows.

What The Study Measured

The study tracked organisations across the Caribbean that completed the board training and followed their results over the months afterward. The data literacy score sits inside a wider set of movements, every one of them from the same study.

Two further findings resist a single number. Communication improved in both directions, bottom-up and top-down, with teams using AI tools to translate and share information. And gender-related bias and equity considerations were built into the training and into how boards then reviewed AI work, so the literacy gain arrived attached to a review discipline rather than on its own.

A Board At 1.8 Reads Everything Second-Hand

The pre-training picture will be familiar to anyone who has presented analysis to a board. Every figure a director sees has been prepared by management, filtered through a committee and formatted for a deck. If the director cannot open the underlying data, the summary is the ceiling of what they can know. Questions become requests, requests join a queue, and by the next quarterly meeting the moment has passed.

Coding was the specific barrier the study identified. Before training, a director who wanted to test a hunch against the data needed someone technical to run the query, and the study links that dependence to an expensive second problem: boards approved vendor pitches they could not question. Vendor costs fell by over 70 percent after training for exactly that reason. Leaders who understood how AI development actually works could weigh what the organisation needed against what it was being sold.

The same gap showed up as overcommitment. Before training, executives and managers took on more AI work than they could deliver. Afterward, the study records them cutting the vanity projects and concentrating on initiatives that generated measurable ROI. In our BI practice we see the untrained version of this regularly: a board approves five dashboards nobody asked for while the one pipeline the finance team needs waits on budget.

The Directors Started Doing The Analysis Themselves

The post-training behaviour is the part of the study we read twice. With coding limitations no longer a barrier, board members ran more advanced analysis themselves, vibe-coded working prototypes, and translated information across functions without waiting for a technical intermediary. A director with a question about churn no longer tables it for the next meeting. They pull the data and answer it, in the meeting.

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

The boards went past analysis. They built custom AI tools in-house on an agents-based approach, and the study credits those tools with improving board cohesion and communication. That detail carries a practical lesson for analytics teams. A board that has built even a small working tool understands what a data product costs to build properly, which changes every conversation about scope, timeline and budget that follows.

Literacy Moved Down Through The Business

The board numbers would matter less if the effect stopped at the boardroom door. It did not. Organisation-wide AI literacy rose from 2.0 to 3.7 on the study's internal index, and the study traces the mechanism: board awareness moved down through business lines, to people managers and then to their teams.

"AI literacy at the top is an enablement story. We measured it trickling down from the boardroom through business lines to people managers, and the whole organisation moved from a 2 to a 3.7," Dunkley says.

Communication moved with it. Teams used AI tools to translate and share information upward, and boards sent clearer questions down. For an analytics function, that is the difference between fielding vague requests and receiving a brief you can actually build against.

A Literate Board Raises The Bar For Analytics

A director who can vibe-code a prototype might look like a threat to the analytics profession. In the studied organisations, production work grew instead. Deployments rose from two to four in eight months. Governance, the least glamorous work in data, went from an 11–15 month argument to a 6 month build, with the training moving data governance to the front of the agenda and reducing overall risk. Time to value fell from around a year to around a month. A prototype proves an idea. Turning it into a system the business can run on requires data engineering, validation, monitoring and governance, and boards that understood the difference commissioned that work faster and with clearer requirements.

The study has limits an analyst should name. It covers organisations that completed the training, so it says nothing about boards that declined it. The deployment figure is a large relative move on a small base, two initiatives becoming four. And the literacy indices are internal measures, built to track direction inside the programme rather than to compare across industries. Read with those limits attached, the claim the study supports is narrower and more useful: in the organisations measured, trained boards demanded more analytics, demanded it sooner, and knew what good looked like when it arrived.

The vendor finding belongs here too. Savings of over 70 percent on vendor costs, running to tens of millions of US dollars across the studied organisations, came from boards that could finally judge vendor claims and buy what the organisation actually needed. For an analytics firm, and we are one, a literate buyer is the best client: sharper briefs, and no budget spent on tools that never leave the shelf.

Next Steps For Boards

The sequence the study suggests is short. Score your board's data literacy honestly before anything else, because a board that has never measured it has no baseline to improve from. Train the board before commissioning the next AI initiative, since every downstream number in the study, from governance speed to vendor spend, moved after the board understood the work. Treat director-built prototypes as briefs for your analytics team rather than as finished systems. And put data governance at the front of the agenda, where the trained boards in the study put it.

Frequently Asked Questions

What did the StarApple AI study find about board data literacy?

In a StarApple AI study of organisations that completed its board-level AI training, board data literacy rose from 1.8 out of 5 to 4 out of 5. Coding limitations stopped being a barrier, so board members could run more advanced analysis themselves, vibe-code working prototypes, and translate information across functions. Organisation-wide AI literacy rose from 2.0 to 3.7 on the same study's internal index, driven by board awareness moving down through business lines to people managers and their teams.

What else changed for organisations whose boards completed the training?

The StarApple AI study recorded AI initiatives reaching deployment rising by more than 50 percent, from two to four over eight months. Time to stand up AI governance and data governance fell from 11–15 months down to 6 months. Vendor costs fell by over 70 percent, with savings across the studied organisations running to tens of millions of US dollars, and time to value fell from around a year to around a month. Communication improved bottom-up and top-down, and gender-related bias and equity considerations were built into how boards reviewed AI work.

Does a vibe-coded board prototype replace an analytics team?

No. In the StarApple AI study, directors' prototypes proved ideas and sharpened briefs, and boards also built custom in-house AI tools on an agents-based approach that improved cohesion and communication. Turning a prototype into a governed, tested production system remains analytics and engineering work, which is why deployments in the studied organisations doubled from two to four in eight months. Literate boards commissioned more analytics work, sooner, and with clearer requirements.

How can a board book StarApple AI's board-level AI training?

Adrian Dunkley, the Caribbean's leading AI expert, has led more than 100 board-level AI training engagements through StarApple AI. Boards can request the full study findings or book a training at starappleai.org or by writing to insights@starapple.ai.

Related reading across the Caribbean AI network