This post is an expanded version of the original essay published on Medium in February 2026. Read the original on Medium ↗
Most writing about the "AI apocalypse" is too vague to act on and too alarming to ignore. It forecasts the end of knowledge work or the collapse of the middle class, then stops without naming who is exposed or what they should do. The exposure is specific, and so is the preparation.
I have built AI systems since 2011, and in 2016 I founded StarApple AI, the Caribbean's first AI company. I have trained more than 500 professionals in AI and data analytics and advised governments in the region on this same question. My view comes from that work, not from a forecast model.
This wave automates thinking tasks, not factory work
The familiar story is robots replacing physical labour. That happened in manufacturing and is happening in warehouses. The current wave is different: large language models write, code, summarise, translate and analyse. They do not do all of these well, but they do enough of them, cheaply enough, to change what an employer will pay a person to do.
The International Labour Organization's 2023 study of generative AI and jobs found clerical work to be the most exposed occupational group, and found that most jobs are more likely to be changed than eliminated. That matches what I see in Caribbean offices.
Who is exposed first
High-volume routine cognitive work. Data entry, standard reports, transcription, scripted customer service, first-pass document review and template writing. AI already does much of this at the level of a junior employee, for a fraction of the cost.
Entry-level jobs in the professions. This worries me most for the region. Junior roles in law, accounting, consulting, journalism and software were where young professionals learned by doing a lot of the routine work. A senior professional with an AI assistant now does much of that work alone. The graduates still exist; the rungs they used to climb are fewer.
Countries with thin infrastructure and slow policy. Places with strong digital infrastructure and training systems will absorb the shift faster. The Caribbean has talented people but uneven broadband, small training budgets and policy that lags the technology.
What the people who adapt have in common
I do not expect mass permanent unemployment. I expect a painful restructuring, and a transition that can do real damage to households who are caught in it. Among the professionals I have trained and the organisations I have advised, the ones who adapt share four traits.
- Deep domain expertise. They can tell when the AI is wrong. AI makes a competent person faster and an incompetent one wrong at greater speed.
- Tolerance for not knowing. No one can track every release. They keep a working picture and update it monthly.
- A learning habit. They pick up a new tool in a week because they have done it many times before.
- Skills that depend on trust. Negotiation, leadership, persuasion and building relationships over years. A model release does not replace a client who trusts you.
Where the Caribbean is most exposed
The sectors that employ the most educated Caribbean workers are business process outsourcing (BPO), financial services, public administration, tourism and education. All of them contain large amounts of routine cognitive work.
BPO is the clearest risk. It employs tens of thousands of people across Jamaica, Trinidad and Tobago, Barbados and the wider region, much of it in voice support and back-office processing, which AI agents now handle for global clients. The region cannot assume its cost advantage will hold when the comparison is with software rather than with another country's wages.
The policy response I would push for has five parts:
- AI literacy built into national workforce programmes, funded as core training rather than a pilot.
- Retraining for BPO workers towards roles that supervise and quality-check AI systems, which the same firms will need.
- Support for local companies that build AI products, so the region earns from AI and does not only buy it.
- School curricula that teach students to use and check AI tools from secondary level.
- Income support for displaced workers during retraining, planned before the layoffs rather than after.
Preparation that pays back, by level
Individuals: use AI tools on your real work every week. Five hours of deliberate use a week for six months will teach you more than any course, because you learn where the tools fail on your tasks.
Organisations: list your ten most time-consuming processes and mark which ones AI could do or assist. Pick one, run a 90-day trial, and measure hours saved and errors caught. An AI strategy is that list plus a sequence, reviewed every quarter, over roughly 18 months.
Governments: many sectors face this within five years, not twenty. Policy written in 2022 predates tools that now write code and run multi-step tasks. National AI strategies need budgets and deadlines, and workforce transition money needs to be allocated before the job losses show up in the labour force survey.
Why I still use the word "apocalypse"
The Greek root, apokalypsis, means an uncovering. This shift is uncovering which skills were valuable only because they took people a long time, which institutions can adapt, and which economies depended on cheap cognitive labour. The weakest part of my argument is timing: I do not know how fast the BPO sector will contract. The preparation above is worth doing at either speed.
What to do next
- If you are employed in a routine office role: this week, pick the task that takes you the most hours and try doing it with an AI assistant. Keep a simple log of time taken and mistakes found for four weeks.
- If you manage a team: within 30 days, list your ten most time-consuming processes, choose one for a 90-day AI trial, and record the baseline hours before you start.
- If you run a BPO operation: identify which of your contracts are mostly scripted voice or data processing and start training supervisors to manage AI-handled queues before clients ask for it.
- If you work in a ministry or training agency: put a line item for AI literacy in the next budget cycle, with a target number of workers trained and a date.
- If you are a student or recent graduate: build a small portfolio of work where you used AI and then checked and corrected it. That shows an employer the judgement they now pay for.
Frequently asked questions
How can I check whether my own job is exposed to AI?
Break your week into tasks and estimate the hours each takes. Tasks that are text-based, repetitive and judged against a clear standard are the most exposed; tasks that rely on relationships, physical presence or accountability for a decision are the least. The ILO's 2023 study on generative AI and jobs publishes exposure scores by occupation that you can compare against.
Where can Jamaicans get subsidised AI or digital skills training?
HEART/NSTA Trust is Jamaica's national training agency and runs certified digital and technical programmes, many free or subsidised. Universities and private providers also run short AI courses. Before paying for one, check whether it includes hands-on work with real tools and whether the certificate is recognised by the employers you are targeting.
Which figures should I watch to see whether AI is cutting Caribbean jobs?
In Jamaica, the Statistical Institute of Jamaica (STATIN) publishes a quarterly Labour Force Survey with employment by industry and occupation. Watch clerical and support roles and the business services industry over several quarters, not a single release. Industry bodies such as the Global Services Association of Jamaica also report BPO employment.
Is it better to learn to code or to learn to use AI tools?
For most workers, learning to use and check AI tools inside their own field pays off faster than learning to code from scratch. Coding still matters for people who want to build AI products, and basic scripting helps anyone automate their own work. The common mistake is taking a general AI course and never applying it to real tasks from your job.
Will AI replace BPO jobs completely in the Caribbean?
Complete replacement is unlikely in the near term, because complex, emotional and regulated calls still need people and clients want human escalation paths. The larger risk is fewer entry-level seats and lower rates for simple work. Firms that retrain agents to supervise AI queues and handle escalations keep more of their contracts.
This is an expanded version of the essay originally published on Medium. Read and share the original here ↗