AI Just Called the Caribbean’s Data Bluff…Again. Here Are 3 Businesses to Build.

Everyone’s busy debating how to share data. How Cute. But here’s the question nobody wants to answer: share what, exactly? You can’t pass around what you never bothered to collect, clean, standardise, and actually interrogate for actionable intelligence. I’ve been saying it for years – the Caribbean is allergic to data. We’d rather guess, vibe, and quote a study from 2011.

Right now, because it looms large, the Caribbean is having yet another conversation about AI.

And, as usual, we’re talking about data sharing.

Recently, the Jamaica Observer reported from the Caribbean Development Bank’s first Caribbean Evaluation Space, where a plenary on “AI, Data, and Power” landed on a familiar diagnosis: the region’s AI ambitions are being held back because countries and institutions aren’t sharing enough data.

Fair.

But I think we’re asking the wrong question.

Because the conversation assumes something that desperately needs challenging:

That the data already exists. It’s just sitting somewhere behind a locked door.

Anyone who has tried to pull comparable data across Caribbean countries knows that’s not quite how this works.

Sometimes the data is locked away.

Sometimes it’s a PDF.

Sometimes it’s in a government filing cabinet.

Sometimes a funded international development agency project stopped in 2019.

Sometimes the company that built that government agency website had a cybersecurity incident and trust levels are in hell.

Sometimes Jamaica, Trinidad, Haiti, and Barbados are measuring the same thing three completely different ways.

And sometimes?

Nobody collected the damn data in the first place.

That’s not a sharing problem.

That’s a pipeline problem.

And until we fix the pipeline, talking about open data and AI is putting the cart so far ahead of the horse that the horse has left the building.

The Caribbean’s data problem is a stack

There are at least four different problems hiding underneath that “we need to share more data” conversation.

And sharing is actually the easiest one.

1. COLLECTION: Is the data being captured at all?

Before you can share data, clean data, analyse data, or train a model on data…

You actually need the data.

This is where things get interesting.

We have decent data on tourism arrivals.

But what about startup formation?

Digital payment volumes?

Creator Economy income?

ICT employment by role?

Remittances by corridor?

Cross-border e-commerce?

AI adoption by Caribbean businesses?

The economic activity is happening.

The data about that activity often isn’t.

And that’s a massive problem if we’re trying to build AI that understands Caribbean economies, consumers, businesses, and behaviour.

This is also the least sexy part of the conversation.

Nobody wants to stand on a stage and talk about registries, sensors, survey design and statistical capacity.

But that’s the work.

You don’t fix data collection with a keynote.

You fix it with people, systems and money that survive the next election.

2. CLEANING + STANDARDISATION: Can the data actually talk to each other?

Let’s say the data exists.

Now comes another Caribbean classic:

Can we actually use it?

Trinidad counts it one way.

Jamaica counts it another.

Barbados has its own definition.

Someone has the data in Excel.

Someone else has it in a PDF.

Someone’s database hasn’t been updated since 2021.

And somewhere, buried inside a government website, is exactly the number you need – except nobody knows where.

This is where interoperability becomes incredibly important.

Because interoperability isn’t just a technology problem.

It’s a “can fifteen governments agree on what this word means?” problem.

That’s much harder.

And there’s a talent problem underneath it too.

The survey cited at the conference found that 60% of Caribbean businesses struggling with digital adoption point to a shortage of skilled IT workers.

Guess what?

Data cleaning, standardisation, integration and infrastructure require skilled people.

You cannot open-source your way around an absent workforce.

3. SHARING: Can anyone actually get to it?

Yes.

This one is real.

Caribbean institutions absolutely have a tendency to treat data like classified intelligence.

Ask for something, and suddenly you’re navigating privacy policies, institutional permissions, “we’ll get back to you,” and a six-month journey to nowhere.

We have absolutely confused data privacy with data secrecy.

And that is costing us.

Researchers can’t research.

Entrepreneurs can’t build.

Investors can’t see.

Governments can’t evaluate.

And AI systems end up learning Caribbean realities from datasets created somewhere else by others with a very different agenda.

So yes, open the damn data.

But here’s the thing:

Open bad data, and you don’t get Caribbean AI.

You get frustrated researchers and a really nice press release.

Sharing matters.

It just isn’t the whole problem.

4. INTELLIGENCE: Is anyone actually turning the data into decisions?

This may be the biggest gap of all.

We are very good at producing reports.

Very good at commissioning studies.

Very good at launching frameworks.

Very good at holding consultations about the frameworks.

And occasionally, very good at commissioning another study about the first study.

But the real value isn’t in the report.

It’s in what somebody does differently because of the data.

That’s the intelligence layer.

A minister changes a policy.

A bank changes its lending model.

An insurer changes its pricing.

A hotel changes its inventory.

A founder spots a market.

An investor sees a sector taking off.

That’s where AI starts becoming economically interesting.

And it requires a different kind of person inside our institutions or outside of them with fresh eyes.

Not just people who can collect and compile.

People who can interrogate the data and turn it into decisions.

That’s why the point from Jos Vaessen of the World Bank’s evaluation group matters: governments need to actively recruit young tech professionals into the public sector.

Because evaluation is becoming an AI problem, whether the public service is ready for it or not.

And here’s where it gets REALLY interesting. Every broken layer of this pipeline is also…a BUSINESS.

The governments aren’t going to build all of this fast enough.

And global vendors aren’t necessarily going to build it for Caribbean-sized markets, nor do we want them owning all of our shit either.

Which means there are some very, very interesting businesses sitting in the gaps.

1. Build the Bloomberg terminal for the Caribbean.

Someone needs to collect the data that nobody else is collecting.

Tourism.

Remittances.

Digital payments.

Startup formation.

Creator economy.

ICT jobs.

E-commerce.

Sector-by-sector, build the datasets.

Keep them updated.

Make them clean.

Make them searchable.

Then sell access through subscriptions and APIs to banks, investors, governments, development agencies, and AI companies.

The moat isn’t some fancy AI model.

The moat is becoming the place everyone goes because you have the data nobody else bothered to collect.

That’s a very different business.

And potentially a very valuable one.

One Startup To Watch: https://dataffluent.com/

2. Build the boring data infrastructure company.

This is probably the least sexy business on the list.

Which is exactly why I like it.

Fifteen governments.

Hundreds of ministries and agencies.

Years of spreadsheets, PDFs, databases and inconsistent definitions.

Someone has to turn that mess into machine-readable, interoperable infrastructure.

A Caribbean company that can walk into a ministry, clean ten years of data, standardise it and make it talk to the system next door?

There is money in that.

Add AI tooling, and suddenly the margins get very interesting.

This is the pick-and-shovel play.

And I suspect the people who build it will make more money than many of the people talking about AI.

Startups To Watch: ExplainingTheCaribbean and FamaCharts

3. Don’t build “AI for the Caribbean.” Build AI for ONE thing.

Please.

We don’t need another “AI platform for the Caribbean.”

Pick a problem.

A painful, expensive, recurring problem.

Hurricane-risk pricing for insurers.

Tourism demand forecasting.

Remittance intelligence for banks and fintechs.

Agricultural yield and pricing.

Creator-economy analytics.

Cross-border payments.

Whatever.

Take one vertical.

Get the data.

Understand the Caribbean context better than a Silicon Valley model ever could.

Then build the thing that turns that information into a decision somebody will happily pay for every month.

That’s the product.

And that’s where the serious valuations could be.

The opportunity isn’t waiting for government to fix this.

Here’s the part I think we should pay more attention to.

None of these businesses require the open-data initiative to pass.

In fact, they exist partly because it hasn’t.

That’s the opportunity.

Don’t wait for government to create the perfect data ecosystem.

Build the company that solves the problem government can’t solve quickly enough.

Then sell it to the banks.

The investors.

The insurers.

The hotels.

The governments.

The global companies trying to enter Caribbean markets.

The AI companies that suddenly realise they have almost no Caribbean training data.

That’s a market. A huge market. And my God, let it come from a Caribbean person wherever they are in the world.

So what should 2027 actually look like?

It cannot simply mean launching another open-data portal and declaring victory.

Please. We’ve done that, and I am sure we’re all tired of that bullshit.

Implementation should mean building the pipeline underneath it.

Collect the data.

Standardise it.

Make it interoperable.

Open what should be open.

Build the talent.

And, most importantly, build the intelligence layer that turns all of that data into decisions, products, and money.

Because here’s the uncomfortable truth:

The Caribbean isn’t data-poor.

We’re data-pipeline poor.

And AI didn’t create that problem; business and government culture did.

And it just made it impossible to keep pretending we don’t have one.

Ingrid Riley is the Founder & Chief Writer of this globally syndicated SiliconCaribe® blog. She is an award-winning entrepreneur who has produced over 380+ Tech Events across 15 Caribbean and Diaspora cities + Media. Communities for 150,000+ people. Platforms for the Global Caribbean Tech & Innovation Economy.

Never miss an event, profile, or story about the global Caribbean tech and Innovation Economy. Subscribe to the free SiliconCaribe Weekly newsletter.