Caribbean Sector Spotlights: How AI Adoption Looks Different in Tourism, Manufacturing, and the Public Sector

 

IN THIS ARTICLE    Reading time: 22 minutes The tenth article of twelve, and the second article of Act IV — The Decision. Article 9 set out an AI governance frame in the abstract — eleven decisions, a Decision-Rights Matrix, two structural questions. Article 10 asks whether the frame holds up against three specific Caribbean sectors: tourism, manufacturing, and the public sector. The point of this article is not to deliver three short case studies. It is to demonstrate that the contextual considerations differ enough across sectors that the same governance frame produces different answers for each — and to give Caribbean directors a tool for locating their own sector against the AI considerations that matter most for it. By the end of this article you will be able to: 1.   Recognise four cross-sector dimensions — regulatory exposure, data-residency intensity, physical-world coupling, and public-legitimacy stake — that, in our engagement experience, determine where AI governance pressure falls hardest in any given Caribbean sector. 2.   Apply the D-AGENTICA™ Sector AI Adoption Profile as the named instrument, locating your institution’s sector against those four dimensions and identifying the two or three considerations that should anchor your AI governance work this year. 3.   Read the three Caribbean sector spotlights — tourism, manufacturing, and the public sector — for the structural lessons each carries, and recognise which lessons translate to your own sector even when the surface details do not.

Three phone calls

Three phone calls, in roughly the same week of this year, from three Caribbean institutions whose AI governance questions sounded superficially similar and proved on closer inspection to be entirely different. The chief executive of a Caribbean tourism group telephoned to ask whether her group’s AI-driven dynamic-pricing engine, which used guest data sourced from a US-based booking platform, was exposing the group to liability under data-protection regimes she had only a glancing knowledge of. The chief operating officer of a Caribbean manufacturer telephoned to ask whether his plant’s predictive-maintenance system, which had been quietly running for eighteen months and which now flagged equipment failures with high accuracy, should be allowed to initiate shutdowns rather than recommend them. The permanent secretary of a Caribbean ministry telephoned to ask how his department should respond to a citizen request — formally lodged, professionally drafted — for the full reasoning behind an AI-assisted decision the department had communicated three weeks earlier.

Three sectors. Three institutions of meaningful scale. Three phone calls in the same week. And three governance problems with almost nothing in common except the word AI in the question. The tourism executive’s problem was data residency and cross-border legal exposure. The manufacturer’s was the boundary between AI-as-advisor and AI-as-actor in a physical context. The civil servant’s was the legitimacy of algorithmic decision-making against a citizen’s right to a human-readable account. Each of these is a real problem; each requires a real answer; the answers are different. This article is about what makes them different and how a Caribbean board chair should think about her own sector against the considerations that have produced the questions in others.

Where we are in the series

Nine articles have come before this one. Acts I through III gave us orientation, guardrails, and application. Act IV opened in Article 9 with the Decision-Rights Matrix — eleven AI decisions mapped across the board, the risk committee, the audit committee, and management. Article 10, this one, asks whether that frame holds when moved from the abstract into specific sectoral contexts. Article 11 will then unveil the comprehensive D-AGENTICA™ Maturity Model — across all domains, with the Finance Function model from Article 8 as its first domain instance. Article 12 closes the series with the call to the Caribbean boardroom that the entire programme has been building toward. We are now, in other words, two articles from the close. The instruments accumulated to this point — nine of them, soon ten — collectively constitute the Dawgen Global position on what good AI governance for the Caribbean should look like.

What this article adds, against everything that has come before, is a contextualising discipline. The risk in any twelve-part series is that the abstract frame becomes the point and the reader is left to translate it back to her own circumstances. Article 10 does that translation work openly, against three concrete sectors, on the premise that a Caribbean director who reads the three sector spotlights below should leave the article better equipped to read the fourth sector — her own — than she was before.

The D-AGENTICA™ Sector AI Adoption Profile

Before the three spotlights, the named instrument. The D-AGENTICA™ Sector AI Adoption Profile is a four-dimension instrument — regulatory exposure, data-residency intensity, physical-world coupling, and public-legitimacy stake — applied at the sector level on a low-medium-high scale. The profile’s purpose is not to score a sector against some absolute benchmark; it is to surface, for any given Caribbean institution, which of the four dimensions are the ones that should anchor the board’s AI governance work in that sector. Different sectors produce different profiles; different profiles imply different governance priorities; the same Decision-Rights Matrix produces different cell entries depending on the profile.

Regulatory exposure asks how heavily an institution in this sector is regulated with respect to AI, either directly (sector-specific AI regulation) or indirectly (financial-services regulation, public-sector procurement law, data-protection law) and whether the regulator is active or merely formal. Data-residency intensity asks how dependent the institution’s AI deployment is on data that crosses jurisdictional boundaries — a tourism group depending on US-hosted booking data is high; a manufacturer running predictive maintenance on its own equipment data is low. Physical-world coupling asks whether the institution’s AI systems control or directly affect physical equipment, infrastructure, or human safety — a manufacturer with predictive-maintenance shutoff authority is high; a tourism group with dynamic-pricing recommendations is low. Public-legitimacy stake asks whether the institution’s AI decisions affect citizens or the public in ways that create a right to challenge — a public-sector entity making benefits decisions is very high; a private manufacturer is low.

THE D-AGENTICA™ SECTOR AI ADOPTION PROFILE
Three Caribbean sectors, four dimensions, scaled from low to very high. The profile surfaces which dimensions should anchor the board’s AI governance work for a given sector. Different profiles imply different governance priorities.
Dimension	Tourism	Manufacturing	Public Sector
Regulatory exposure	Medium	Low	High
Data-residency intensity	High	Low	Medium
Physical-world coupling	Low	High	Low
Public-legitimacy stake	Low	Low	Very High

 Low   the dimension is not a primary governance driver in this sector      Medium   the dimension warrants explicit board attention      High   the dimension is a top-three governance priority      Very High   the dimension is a defining sectoral consideration

The three spotlights below illustrate the profile in action. Each spotlight opens with a one-line profile statement, then walks through the same four-element analysis: regulatory pressure, adoption pattern observed, distinctive risk, and governance implication. The reader who absorbs the three spotlights should be able to write a fourth — for her own sector — without further instruction.

Spotlight 1 — Caribbean Tourism

Profile: Regulatory exposure medium. Data-residency intensity high. Physical-world coupling low. Public-legitimacy stake low.

Regulatory pressure

Caribbean tourism is, on the face of it, lightly regulated with respect to AI. There is no Caribbean tourism-specific AI regulation. The sector is governed by the same general data-protection regimes that apply across the region — Jamaica’s Data Protection Act of 2020, similar instruments in Trinidad and Tobago, Barbados, the Bahamas, and the OECS — and by the consumer-protection law of the visiting guest’s home jurisdiction. The latter is where the regulatory pressure actually lands. A Caribbean hotel group serving European guests is subject to the EU General Data Protection Regulation in respect of those guests’ data, regardless of where the data is processed. A group serving US guests faces the increasingly stringent landscape of state-level US privacy law. The sector’s regulatory exposure is, in other words, imported — and Caribbean operators sometimes underestimate it because the regulator is not in the Caribbean.

Adoption pattern

Tourism is one of the fastest-adopting Caribbean sectors. Dynamic pricing engines, guest-segmentation models, demand-forecasting tools, AI-assisted concierge agents, and review-aggregation models are all in production at multiple Caribbean tourism groups we work with. Most of these are vendor-supplied; few are home-grown. The adoption pattern is therefore a vendor concentration pattern — a small number of international booking and revenue-management platforms supply AI capability to a large number of Caribbean operators, with the data flowing to platforms whose primary jurisdictions of registration are the US, Western Europe, and increasingly Singapore.

The distinctive risk

The distinctive risk for Caribbean tourism is data-residency exposure under imported regulatory regimes. A Caribbean hotel group whose AI engine processes guest data on a US-hosted platform is, simultaneously: subject to Jamaican data law for the data it owns; subject to US state law because the platform sits there; and subject to the EU GDPR for any guest who is an EU resident. The combination produces a multi-jurisdictional compliance question that no single legal team in the region is fully resourced to handle, and that the international platforms tend to disclaim in their terms of service. We have seen Caribbean hotel groups assume their vendors carry the regulatory risk and discover, in the course of an EU regulatory inquiry, that the contract had assigned the risk back to them.

Governance implication

The Decision-Rights Matrix from Article 9 produces, for a Caribbean tourism group, a specific emphasis. Decision 6 — data architecture — moves up in importance because the data-residency question is structural. Decision 7 — AI risk appetite — must include explicit treatment of foreign-jurisdiction regulatory exposure, with the board owning the position rather than allowing management to inherit it from vendor terms of service. Decision 10 — disclosure — becomes more material because what the tourism group says about its AI to its guests is now a regulated communication in the guests’ home jurisdictions. The board’s emphasis, in other words, shifts toward the data-and-disclosure rows of the matrix; the operational rows (use cases, vendor selection) can responsibly remain at risk-committee or management level.

WHAT WE OBSERVE ACROSS CARIBBEAN TOURISM ENGAGEMENTS In the Caribbean tourism engagements we have undertaken in the past twelve months, the most consistent finding is the gap between the institution’s view of where its guest data resides and where it actually resides. In four engagements out of five, the operator believed the data was in a single jurisdiction; in fact, it was distributed across two or three depending on the specific platform feature in use. This is not a vendor failing; it is a disclosure-and-comprehension failing on the buyer side. The remedy is a documented data-flow map maintained at board level and refreshed annually — a piece of work most Caribbean tourism boards have not yet commissioned and would benefit from doing this year.

Spotlight 2 — Caribbean Manufacturing

Profile: Regulatory exposure low. Data-residency intensity low. Physical-world coupling high. Public-legitimacy stake low.

Regulatory pressure

Caribbean manufacturing operates in a regulatory environment that is, by comparison with tourism, structurally light on AI-specific obligation. The sector is regulated for occupational safety, environmental compliance, and product standards, but no Caribbean manufacturing regulator has yet issued AI-specific guidance. The regulatory pressure that does exist is adjacent — through occupational safety and health regulation when AI controls equipment, and through international product certification regimes when AI is part of the manufacturing process for an exported product. The sector’s regulatory exposure is real but indirect; it bites where the AI touches the physical world.

Adoption pattern

Caribbean manufacturing has been slower to public AI adoption than tourism but, in our experience, more mature in substance where adoption has occurred. Predictive-maintenance models, computer-vision quality-control systems, supply-chain demand-forecasting, and energy-optimisation models are all in production at the larger Caribbean manufacturing groups, often with engineering rather than IT ownership and often built on the institution’s own equipment data rather than licensed from external vendors. The adoption pattern is therefore engineering-led, data-internal, and physical-world-coupled. This is a meaningfully different shape of AI adoption from the vendor-driven, data-imported pattern in tourism.

The distinctive risk

The distinctive risk for Caribbean manufacturing is the boundary between AI as advisor and AI as actor in a physical context. A predictive-maintenance model that flags equipment for inspection is one thing; the same model with authority to initiate a shutdown is a different thing — and the controls discipline for the second is genuinely different from the first. When an AI system can take a physical action, the governance questions that arise are not the controls questions familiar from financial AI (Article 8) but operational-safety questions that draw on a different professional discipline entirely. We have observed Caribbean manufacturing boards, advised by IT-led AI assessments, fail to surface the engineering-safety dimensions of their own AI deployment because the assessor was not an engineer.

The risk that follows is asymmetric. A financial AI model that fails produces a financial-statement misstatement, which is recoverable. A physical-world AI system that fails — that initiates the wrong shutdown, or fails to initiate a needed one — can produce equipment damage, production loss, or, in the worst cases, injury. The downside is not merely larger; it is categorically different, and the governance question is whether the board has had a serious conversation about where on the advisor-to-actor spectrum its physical-world AI is, and where it should be.

Governance implication

The Decision-Rights Matrix produces, for a Caribbean manufacturer, a different emphasis from tourism. Decision 8 — control framework — becomes the most important row, with the risk committee owning a specific question: at what point on the advisor-to-actor spectrum does an AI system require a parallel safety case, independent of the AI engineering team? Decision 11 — crisis response — takes on physical-world dimensions that boards in services-sector contexts can responsibly leave at management level but boards in manufacturing cannot. The risk-appetite decision (decision 7) must explicitly cover physical-world AI tolerances. And the single most important sectoral move a manufacturing board can make is to ensure that the AI risk owner reports to a director with engineering experience, not to one with IT or finance experience alone. This is a small organisational decision that has meaningful governance consequences.

Spotlight 3 — Caribbean Public Sector

Profile: Regulatory exposure high (and rising). Data-residency intensity medium. Physical-world coupling low. Public-legitimacy stake very high.

Regulatory pressure

The Caribbean public sector is, of the three sectors examined here, the one whose regulatory exposure is rising most rapidly. Public-sector AI is governed by administrative-law principles that long predate AI — natural justice, procedural fairness, the duty to give reasons — and these principles apply to AI-assisted decisions as a matter of constitutional law in most Caribbean jurisdictions. To this structural exposure is being added an increasingly explicit AI-specific overlay: data-protection regulation has matured across the region, freedom-of-information regimes are increasingly used to test AI decisions, and judicial review applications challenging algorithmic determinations are appearing in Caribbean courts. The sector is also subject to the procurement disciplines that govern public-sector contracting, which are now beginning to incorporate AI-specific provisions in several jurisdictions.

Adoption pattern

Caribbean public-sector AI adoption is uneven. Some ministries — typically those responsible for revenue administration, customs, social benefits, and certain regulatory determinations — have been substantial adopters of classical machine learning for fraud detection, eligibility scoring, and risk-based audit selection. Other ministries have not yet begun. The adoption that has occurred has typically been technical-team-led with senior policy leadership informed but not closely engaged. The pattern, in other words, is technical-internal, with AI serving administrative determinations that affect citizens directly.

The distinctive risk

The distinctive risk for the Caribbean public sector is the legitimacy of algorithmic decision-making against a citizen’s right to a human-readable account. When a private-sector AI declines a credit application, the customer’s recourse is to take her business elsewhere; when a public-sector AI declines a benefits application, the citizen’s recourse is judicial review, and the courts will not accept “the model decided” as an answer. The duty to give reasons that a public official traditionally discharges by writing a memorandum cannot be discharged by AI alone; the AI’s output is, at best, a contribution to a reasoned decision that a human official remains responsible for. Several Caribbean jurisdictions have already produced administrative-law decisions reflecting this, and more are expected. The risk for a Caribbean ministry deploying AI without having thought through the reasons-giving obligation is not a regulatory fine; it is judicial review of the underlying determinations and, in serious cases, the setting aside of every determination the AI assisted.

Governance implication

The Decision-Rights Matrix produces, for a Caribbean public-sector entity, a third distinctive emphasis. Decision 1 — strategic posture on AI — becomes a public-policy decision that should sit not only with the entity’s senior leadership but with the responsible minister, with appropriate parliamentary engagement where the AI use is material. Decision 4 — specific use cases — must be reviewed against the administrative-law standard of procedural fairness before deployment, not after a first determination is challenged. Decision 10 — disclosure — takes on the constitutional dimension of the duty to give reasons; an AI-assisted determination must be communicable to the citizen in human-readable form, with the human official’s reasoning visible above the AI’s contribution. The most important sectoral move for a public-sector entity is to require that every AI-assisted determination produce a human-readable rationale that the responsible official has personally adopted; the AI is treated as input, not decision-maker.

WHAT WE OBSERVE ACROSS CARIBBEAN PUBLIC-SECTOR ENGAGEMENTS In the Caribbean public-sector AI engagements Dawgen Global has been involved in over the past two years, the gap most often identified is between the technical adoption of AI within an administrative function and the procedural-fairness review of the specific use case. Technical teams build AI systems that perform well on the metric they were designed against; the administrative-law dimension of whether the system supports a procedurally fair determination is rarely surfaced before deployment. We have not yet seen a Caribbean ministry where this question was treated as a pre-deployment gate; we have seen it become an issue post-deployment in three engagements. The gate is straightforward to install; the cost of installing it after the fact is materially higher than installing it in advance.

What the three spotlights teach a director in a fourth sector

The three spotlights together carry four lessons that translate to any Caribbean sector, and that a board chair in financial services, healthcare, telecommunications, agribusiness, energy, retail, real estate, or any other sector should be able to apply directly to her own context.

First, the regulatory pressure on a Caribbean institution’s AI deployment is rarely the regulator the institution most often deals with; it is the indirect regulator — the foreign data-protection authority, the international product-certification body, the administrative court reviewing an algorithmic determination, the procurement regime that governs public contracting. The board chair’s first task is to identify which indirect regulators have the strongest claim on her institution’s AI use, and to ensure her risk committee is briefed on those, not just on the local regulators it knows.

Second, the data-residency question is the most underexamined dimension of Caribbean AI governance and the one most likely to produce surprises. Wherever AI capability is supplied by an international vendor — which is most cases — the data is almost certainly crossing jurisdictions in ways the buyer institution does not fully understand. A documented data-flow map maintained at board level is the single most useful institutional artefact a Caribbean board can commission this year, and is the artefact most often missing in our engagement experience.

Third, where AI couples to the physical world, the governance question moves outside the IT-and-finance professional disciplines and into engineering safety. Boards in any sector with physical-world coupling — manufacturing, energy, water, transport, mining, agriculture — should ensure their AI risk owner reports to a director who understands operational safety, not solely to one who understands technology or finance.

Fourth, where AI affects citizens or members of the public in ways that create a right to challenge — public sector, regulated utilities, healthcare, certain financial services — the AI is treated as input to a human decision, not as the decision itself, and a human-readable rationale is preserved for every determination. The institutional cost of doing this in advance is small. The cost of doing it after a court has ordered it is large.

This fourth lesson is worth dwelling on because it applies to many more Caribbean institutions than the public-sector spotlight alone might suggest. Any Caribbean institution making determinations that materially affect the legal rights, property, livelihood, or access-to-essential-services of a citizen or customer operates within the legitimacy framework the public-sector spotlight described — even where the institution is private and the framework is contractual rather than constitutional. A bank declining a mortgage, an insurer declining a claim, a telecommunications provider terminating service, a healthcare provider denying a procedure: each of these decisions, when AI-assisted, generates the same reasons-giving expectation that a public-sector benefits decision generates, and the institutional discipline of a human-readable rationale produced by an accountable human decision-maker should apply. We have observed Caribbean financial institutions begin to deploy AI in credit and claims contexts without explicit reasons-giving discipline, and the issue has surfaced in regulatory examination conversations sooner than the institutions expected. The cost of installing the discipline before it is required externally is trivial; the cost of installing it after the fact, retroactively, against a determinations log of months or years, is not.

What the board chair should do this year

The article has now covered ground enough to be specific about what a Caribbean board chair should commit to in the next ninety days as a sectoral overlay to the Article 9 governance work. Three steps, in order.

First, the chair should commission a one-page application of the D-AGENTICA™ Sector AI Adoption Profile to her institution’s specific sector and circumstance — low-medium-high against each of the four dimensions, with a one-sentence justification for each entry. This page becomes the sectoral context document for the board’s AI governance work and is reviewed annually.

Second, the chair should identify, with the company secretary, the indirect regulators whose claim on the institution’s AI use is most material, and ensure the risk committee is briefed on the relevant law of each. This is rarely a long list — for most institutions it is two or three regimes — but the list is almost never written down before it is asked for.

Third, the chair should commission, through the risk committee, a documented data-flow map for the institution’s AI systems showing which data crosses which jurisdictional boundaries. The map need not be exhaustive in technical detail. It should be sufficient to answer, for any AI use case, the question: whose data, where, under whose legal regime? In our engagement experience, this map exists in fewer than one in five Caribbean institutions of scale, and is the single artefact most likely to be requested by an external auditor or regulator in 2026.

By the end of 2026, a board chair who has commissioned this work will have a sectoral overlay to the Decision-Rights Matrix that reflects the specific contours of her institution’s sector, a written list of the indirect regulators her risk committee monitors, and a data-flow map showing which jurisdictions her AI deployment touches. That is a different starting position from the one most Caribbean boards occupy as of writing. It is, again, achievable in two board cycles.

Closing reflection — and what comes next

Article 10 has demonstrated, against three concrete sectors, that the abstract governance frame from Article 9 produces materially different specific answers depending on the sectoral context. The same matrix; different cell weights; different governance priorities; different artefacts to commission. The named instrument introduced in this article — the Sector AI Adoption Profile — is the lens through which a Caribbean board chair locates her institution against the considerations that matter for her sector.

Article 11 will now unveil the comprehensive D-AGENTICA™ Maturity Model — the full multi-domain instrument of which the Finance Function model from Article 8 was the first instance, and against which any Caribbean institution can locate itself across the full range of AI governance and adoption considerations. The model is the longest-developed of the series’ instruments and we have held it back to its proper place in the architecture. Article 12 will then close the series with the call to the Caribbean boardroom that the entire programme has been building toward. Two articles remain. We hope you will be with us for both.

FOR THE BOARD AGENDA This article has specified, through the Sector AI Adoption Profile and the three spotlights, what a Caribbean board chair should expect of herself and her risk committee on the sectoral context of AI governance this year. A board chair, lead independent director, or risk committee chair reading this article has earned the right to ask their leadership team one specific question and to propose one specific decision that will materially improve the position over the next ninety days. THE QUESTION Within ninety days, can the company secretary, working with the chief executive and the chief risk officer, present to the board a completed D-AGENTICA™ Sector AI Adoption Profile for this institution — low-medium-high across regulatory exposure, data-residency intensity, physical-world coupling, and public-legitimacy stake, with a one-sentence justification for each entry — together with a written list of the indirect regulators most material to our AI use, and a data-flow map showing which jurisdictions our AI deployment touches? THE DECISION That, by the end of the next quarter, the institution’s AI Governance Decision-Rights Matrix appendix to the board’s terms of reference will be supplemented with a sectoral overlay reflecting the Profile, the indirect regulators list, and the data-flow map; and that these three artefacts will be reviewed annually thereafter as part of the standing AI governance cycle.

ABOUT THE AUTHOR

Dr. Dawkins Brown is the Executive Chairman and Founder of Dawgen Global. He holds a PhD and the MCMI and ACFE designations, with twenty-three-plus years of professional experience including a prior career at Ernst & Young before founding Dawgen Global. He writes the LinkedIn newsletter Caribbean Boardroom Perspectives and serves as Executive Chairman of Business Access Television.

About Dawgen Global