- Credit Garden's parent, Maestro AI Labs, runs a free international credit calculator at worldcredscore.com, scoring individuals directly on a 0-850 scale across 52 named countries and territories, with the company stating the underlying model is calibrated for 180 in total.
- The score weights payment history at 35%, credit utilisation at 30%, credit history length at 15%, credit mix at 10%, and new credit inquiries at 10%, layered with a country-specific calibration built from GDP per capita, minimum wage, unemployment, inflation, financial inclusion, and banking infrastructure.
- This runs alongside, not instead of, Credit Garden's business-to-business product, which entered beta with three Caribbean credit providers in Q2 2026 and generated preliminary approval-rate gains of 18 to 34%, per Maestro AI Labs' own H1 2026 update, with general availability targeted for Q3 2026.
- The calculator states plainly it is "not an official credit bureau score." It is not a revenue line on its own. What it is: a near-zero-cost channel for exactly the kind of labelled outcome data an emerging-market credit model runs on, collected before the institutional product has even shipped.
- Founder Adrian Dunkley's own bio on the site claims a lending-desk origin for the product: more than 25 years across CIBC FirstCaribbean, the Development Bank of Jamaica, Mayberry Investments, NCB Financial Group, and Barita Investments, including an AI underwriting system he says approved billions of dollars in unsecured loans at Jamaica's largest commercial bank.
A Calculator Nobody Announced
There was no press release. worldcredscore.com is not linked from Credit Garden's own product page on this site, and it does not appear in any prior Maestro AI Labs blog post. It shows up only as a quiet line in the footer of several recent articles here, under "Caribbean AI Network," next to StarApple AI and the Caribbean AI Association. Visit it directly and the pitch is blunt: "World's First International Credit Score." "Know Your Credit Strength. Plan for Freedom." A visitor enters personal details, a credit history, and a country, and receives a score between 0 and 850, sorted into Poor, Fair, Good, and Excellent, the same four-tier language North American bureaus use, applied to a population most North American bureaus have never scored.
The site is explicit about what the number is not. Its own disclaimer states the score "is not an official credit bureau score and should not be used as the sole basis for financial decisions," and points users toward a licensed local bureau for anything that actually needs to be official. That single sentence does most of the work of explaining why this product exists. A tool that cannot be used to approve or deny a loan on its own is not a lending product. It is a lead magnet with a data-collection form attached, and the form asks for exactly the inputs a credit model needs to get better.
The Weights Behind the Score
The methodology page is more detailed than most consumer credit tools bother to publish. Five factors make up the core score, and a sixth layer recalibrates the result to the country selected.
| Factor | Weight | What it measures |
|---|---|---|
| Payment history | 35% | Track record of on-time payments across credit obligations |
| Credit utilisation | 30% | Share of available credit currently in use |
| Credit history length | 15% | How long credit accounts have been active |
| Credit mix and depth | 10% | Variety of credit types held |
| New credit and inquiries | 10% | Recent applications and new account activity |
Those five weights, on their own, would describe almost any FICO-style scorecard built anywhere in the world. The distinguishing layer sits on top: a country calibration factor drawn from GDP per capita, minimum wage, unemployment rate, inflation, a financial inclusion index, and banking infrastructure. In practice, this means the same payment record from a small-business owner in Kingston and one in Toronto is not meant to land on the same number. The model is trying to answer a different question than a traditional bureau does: not "does this look like a good North American borrower," but "is this a good borrower relative to what is achievable in this specific economy." That is the same design principle already stated on Credit Garden's product page here, that the model was "trained on Caribbean, West African, and South Asian financial patterns from the ground up," applied now to a consumer-facing tool rather than a bank's back office.
The Coverage Claim, Checked
Two different numbers describe the calculator's reach, and they are not the same number. The homepage states, in large type, that the World Credit Score is "calibrated for 180 countries." Open the calculator itself, and the country dropdown lists a fixed set of options: by our count, 52 countries and territories named individually, from Jamaica, Trinidad and Tobago, Barbados, the Bahamas, Guyana, and Haiti through most of the English and Spanish-speaking Caribbean, into Mexico, Brazil, Colombia, and a dozen more Latin American markets, then a smaller set spanning the United States, Canada, the United Kingdom, four Western European countries, Nigeria, Ghana, Kenya, South Africa, India, the Philippines, Indonesia, Singapore, and Australia.
That gap between 52 and 180 is not necessarily a contradiction. A model can be calibrated with country-level economic parameters for a much longer list of nations than the number of countries a consumer product currently exposes through a dropdown menu; building the underlying calibration table and shipping a front-end selector for every entry in it are different amounts of work. But an investor doing diligence should treat "calibrated for 180 countries" and "selectable in the live product" as two separate claims until the company shows the difference, not one number restated twice.
A free tool that states outright it is not an official score will never be the revenue line in this business. Read correctly, it is not trying to be. It is trying to be the cheapest way to collect a labelled outcome from a person the institutional product cannot reach yet.
Two Credit Gardens, Running at Once
Maestro AI Labs' own H1 2026 deployment update, published on this site in June, described a different Credit Garden: an institutional product in beta with three Caribbean credit providers since the second quarter, using mobile payment patterns, utility payment history, informal credit obligations, and small business revenue signals to reassess applicants that traditional bureau-based models had declined. The company reported preliminary approval-rate improvements of 18 to 34 percent for those declined segments, with no statistically significant increase in default rates, and set a target of general availability for Caribbean credit providers in Q3 2026, the quarter that ends this week.
Nothing published anywhere confirms that general-availability date has been hit. What is confirmed, because it is live on the public internet right now, is the consumer calculator running in parallel to that institutional pipeline. The two products serve different customers and, on the surface, different purposes: one sells a scoring API to a lender, the other gives an individual a free number. But they draw from the same underlying model family and the same founding thesis, that standard credit bureaus miss the financial behaviour that actually predicts repayment in a thin-file economy. Running both at once, on the same brand, before the institutional side has confirmed it shipped, is a specific kind of bet: that a direct-to-consumer funnel can generate useful signal and market awareness cheaply, on a timeline that does not wait for enterprise sales cycles to close.
What the Caribbean Numbers Actually Say
Credit Garden's own materials put the stakes in blunt terms. The company states that over 70 percent of Caribbean adults are underserved by traditional credit systems, and that more than half of Jamaicans specifically remain outside formal credit. Those are Credit Garden's own figures, not an independent audit, but they sit consistently with the wider $380 billion annual lending gap and 1.7 billion credit-invisible population this site's Credit Garden product page already cites, and with the roughly 18 billion US dollars Caribbean households receive in remittances each year that this site's own reporting has already argued no formal credit bureau scores at all.
That the founder brings a banking career rather than a purely academic one to this specific problem is not incidental. Adrian Dunkley's own biography on Credit Garden's site describes more than 25 years across finance and artificial intelligence, including executive roles at CIBC FirstCaribbean, the Development Bank of Jamaica, Mayberry Investments, NCB Financial Group, and Barita Investments. He states he personally designed an AI underwriting system that approved billions of dollars in unsecured loans while at Jamaica's largest commercial bank. Whatever weight an investor gives a founder's own account of his career, the detail is consistent with the product's design choice: a scoring model built by someone who has sat on the approval side of a Caribbean loan desk is more likely to know which informal signals a bureau throws away actually predict repayment, and less likely to treat "no bureau file" as synonymous with "no data."
What This Is Worth, and What It Is Not
Investors should hold two facts side by side without letting either one cancel the other out. First: a free calculator that explicitly disclaims itself as unofficial has no direct revenue model, and nothing on worldcredscore.com states how, or whether, the data a visitor submits actually trains the institutional Credit Garden product. Without a published data-use policy connecting the two, the consumer tool is, at minimum, brand-building and lead generation for the "comprehensive credit health review" the site also sells through its founder's advisory team, and at most, a genuine acquisition funnel for training data. Diligence on this point means asking the company directly, not assuming the more flattering interpretation.
Second: shipping a working, publicly usable product across 52 countries, this cleanly built, this soon after Maestro AI Labs' own 6 August 2026 public launch, is itself a signal worth weighing. Plenty of early-stage companies promise a consumer-facing wedge to a B2B product and never build it. This one exists, is live, and works. The open questions are not whether the calculator is real. It is whether the company converts the traffic it generates into consented, usable training data, whether it can hold that consent and data-residency posture across 52 jurisdictions with materially different privacy law (a subject this site has covered in detail for Caribbean regulated entities generally), and whether the institutional GA date that quarter is closing on actually lands.
None of that makes the consumer calculator a marketing stunt. A stunt would have been a splashy launch with no methodology page and no disclaimer. What Credit Garden built instead is a quiet, technically serious tool that happens to sit exactly where the company needs more of the one asset its entire pitch depends on: real credit behaviour from people no other model has scored. The strategy is legible. Whether it pays off is a diligence question, not a marketing one, and it is the right one to be asking now, before the institutional side confirms whether it hit its own Q3 target.
For the wider network: StarApple AI, the first AI company established in the Caribbean, remains the parent lineage behind this thesis, and Adrian Dunkley's own site carries the fuller founder history behind Credit Garden's design choices.