Is Caribbean BPO doomed by AI?

The Island AI Brief · Workforce Question

The Caribbean's largest private-sector employment story is under siege. Jamaica's BPO sector has lost between 12,000 and 20,000 jobs from its 2023 peak. The industry says AI is not the main cause. The AI industry says it will be. Both are partly right, and the honest answer decides what happens to 40,000 Caribbean workers this decade.

By the Caribbean AI Newsletter 12 August 2026 12 min read

The Caribbean's largest private-sector employment story is under siege. Jamaica's BPO sector has lost between 12,000 and 20,000 jobs from its 2023 peak of 60,000. The question is whether AI is the cause, whether it will finish the job over the next decade, or whether the sector's real transformation is something different that AI is only accelerating.

Where the numbers stand

Jamaica is the Caribbean's clearest test case. The country hosts the region's largest BPO industry, employing an estimated 40,000 to 50,000 people as of March 2026 (the higher figure comes from the government, the lower from industry association GSAJ). At its 2023-24 peak, the sector employed roughly 60,000 people. That means Jamaica's BPO sector has contracted by between 17% and 33% depending on which counting method one accepts, according to reporting in the Jamaica Observer and Nearshore Americas.

The dollar picture is equally sharp. Jamaica's global digital services sector spent an estimated US$780 million in the local economy in the fiscal year ended March 2026, down US$220 million from about US$1 billion a year earlier, per Jamaica Observer analysis of official figures. That spending still exceeded Jamaica's US$612 million in bauxite and alumina exports for 2025. The sector is smaller than it was, and still one of the country's largest sources of formal employment and foreign-exchange earnings.

60,000
Jamaica BPO peak employment (2023-24)
40k-50k
March 2026 estimate (GSAJ / Government)
US$780m
Local spending, FY ending March 2026
US$220m
Drop in local spending vs prior year
Exhibit 1 · Jamaica BPO employment
The trajectory: growth to 2023-24, then a fall of 12,000 to 20,000 jobs
Government and industry figures diverge on the exact number, but both show a peak and a fall. Both also predate the sharpest AI enterprise CX savings announcements of mid-2026.
70,000 60,000 45,000 30,000 15,000 0 2015 12.5k 2018 25k 2020 36k 2022 50k 2023-24 60k PEAK Mar 2025 55k (govt) Govt 50k GSAJ 40k Mar 2026 Loss: 10,000 to 20,000 jobs Sources: Jamaica Observer, Nearshore Americas, GSAJ, Government of Jamaica reporting (2015-2026)
The peak of 60,000 was reached during a post-pandemic nearshoring surge. The subsequent fall to 40,000 to 50,000 predates the sharpest 2026 enterprise AI announcements, meaning the AI displacement thesis has yet to fully test itself against Caribbean BPO.

The case that BPO is doomed

The 2026 pattern of enterprise headcount announcements makes this case direct. Salesforce CEO Marc Benioff publicly disclosed reducing the company's customer support team from 9,000 to 5,000, a 44% cut attributed to AI agents. HSBC announced up to 20,000 cuts under an AI overhaul, with CEO Georges Elhedery telling staff "don't fight AI." Dow, HP, BT, Verizon, Oracle, and Monday.com all attributed 2026 headcount reductions to AI. Founder Reports counted more than 100,000 technology-sector jobs cut in 2026 with AI cited as a driver.

Industry analysts project that AI could affect up to 50% of customer service roles globally by 2030, with outsourced delivery markets facing the steepest displacement pressure. Offshore voice and text-based operations are the most susceptible to automation at scale. That is a description of Caribbean BPO's core work.

The economics compound the pressure. A Caribbean BPO agent typically earns US$3 to US$6 per hour in wages plus benefits and overhead. A modern AI voice agent handles a customer conversation for roughly US$0.05 to US$0.10 of compute cost. Even accounting for enterprise software fees and human oversight, the cost-per-query gap is roughly two orders of magnitude. Gartner projects that conversational AI will reduce global contact-centre labour costs by US$80 billion in 2026 alone.

On this reading, the Jamaica losses so far are the early tremor rather than the earthquake. The 2023-to-2026 contraction happened before the biggest AI enterprise deployments landed. When Fortune 500 CX buyers sit down at 2027 contract renewals with the Salesforce, HSBC, and Klarna savings benchmarks in hand, offshore voice contracts become where the negotiation starts. Any Caribbean nation whose growth strategy still targets a return to 60,000 BPO seats is designing policy around a market that buyers are already restructuring at contract renewal.

The case that BPO is not doomed

The counterweight begins with Klarna. The Swedish fintech's 2024 announcement that AI had replaced 700 customer service agents became the emblem of white-collar displacement. By early 2026, Klarna had reversed course. Digital Applied and Perspective AI both reported that Klarna was rebuilding human customer service capacity after complex interactions produced customer satisfaction drops that overall volume metrics had masked. The case now sits in every enterprise CX strategy conversation as the reason not to over-index on pure-AI replacement.

Gartner's own projection, cited above, contains a second number. Even as AI reduces contact-centre labour costs by US$80 billion in 2026, only about one in ten agent interactions will be automated. The savings come from AI absorbing the high-volume repetitive tier that has been driving 30% to 45% annual agent turnover across the industry, according to Retell AI's analysis. AI is absorbing tasks inside contact centres rather than shutting them down.

The market is also growing, not shrinking. Grand View Research projects the global BPO market at US$358.6 billion in 2026 and US$695.8 billion by 2033. The composition is shifting toward higher-complexity work in healthcare, financial services, insurance, and regulated industries where hallucination risk, compliance requirements, and empathy make human agents structurally necessary. Buyers are increasingly evaluating providers on their technology stack, security posture, and workforce skills rather than on hourly wage rate.

Jamaica's own industry leadership makes this case explicitly. GSAJ President Yoni Epstein has publicly attributed the sector's losses to low productivity, Hurricane Melissa's October 2025 disruption, a tight labour market, and top-tier companies relocating for stronger incentive packages, not to AI. His predecessor Wayne Sinclair went further, telling the Financial Gleaner that "the adoption of AI may help rather than hurt the sector." itel, Jamaica's largest homegrown CX provider, is investing in AI-driven training for its agents rather than replacing them.

The Klarna lesson, and what it hides

The Klarna story is being read across the industry as evidence that AI cannot replace humans. That reading is too comfortable.

Klarna's original 2024 deployment ran on a tightly scoped consumer-fintech use case with structured data, authenticated users, and a finite set of common intents. It was, as Perspective AI's analysis put it, almost the easiest possible support workload for AI to absorb. On that specific workload, the AI succeeded in absorbing roughly two-thirds of conversations. The reversal happened because the remaining third produced customer satisfaction losses on the complex tier that were masked by good aggregate metrics.

Klarna proved that AI can absorb the easy tier at scale, and that pure-AI replacement without a strong human escalation path costs enterprises more than the initial headline savings suggested. For Caribbean BPO, the operational implication is direct. The easy tier is where price collapses. The complex tier, where regulated industries need audit trails and empathetic conversations require human accountability, is where providers still hold pricing power.

Exhibit 2 · What AI absorbs vs what stays with humans
The two tiers of BPO work, and where the profitable one sits

Where AI already wins on cost

  • Password resets and account lookups
  • Order status and delivery tracking
  • FAQ responses and policy summaries
  • Appointment scheduling and confirmations
  • Basic technical troubleshooting scripts
  • Structured consumer-fintech support (Klarna case)
  • Call routing and intake summarisation
  • Post-interaction summaries and QA notes

Where humans still win on value

  • Insurance claims disputes and adjustments
  • Healthcare member services and clinical intake
  • Financial services escalations and retention
  • Regulated compliance interactions with audit trails
  • Complex empathetic conversations (bereavement, fraud victims)
  • Multi-party negotiations and account resolutions
  • High-lifetime-value customer retention
  • AI oversight, evaluation, and human-in-the-loop workflows

The case that both are half right

The honest answer combines the first two cases. The version of Caribbean BPO built around labour arbitrage, in which young workers in headsets handle password resets at US$4 an hour, is going. That work is exactly where AI voice agents are strongest, where enterprise buyers have the clearest case to switch, and where cost-per-query economics are most brutal. This is what the enterprise 2026 announcements are actually describing.

The version of Caribbean BPO built around English-speaking agents handling regulated, empathetic, high-stakes interactions is not going. That work is where AI has stumbled, where regulated industries require human accountability, and where the value delivered is measured in customer lifetime value rather than conversation count. Jamaica's positioning in this segment is materially different from India's or the Philippines' because of proximity to the US market, cultural affinity, and shared business hours with US buyers. Global buyers pay a premium for those attributes on complex work.

How Caribbean BPO handles the transition between these two versions decides the size, wages, and standing of the sector by 2030. It requires a workforce that can handle KPO-level complexity, a training pipeline that treats AI as a working tool rather than a threat, and government policy that rewards the shift rather than defending the version that is going. Jamaica's BPO recovery task force explicitly targets a shift from basic BPO to higher-value knowledge process outsourcing, and companies including Annotera are pitching Jamaica as an AI data annotation, human-in-the-loop, and model evaluation hub. This transition is a rebuild, and if the region cannot get through it, the raw job loss will fall on the 40,000 workers still employed in the sector.

Where the argument gets harder

Three uncomfortable facts sit inside the third case. First, the transition to higher-value work does not preserve the same number of jobs. A KPO workforce serving complex regulated industries is smaller and more expensive than a BPO workforce answering password resets. Even a successful transition may leave the Caribbean sector at 25,000 to 35,000 jobs rather than 60,000. The workers who do not make the transition still have households to support.

Second, the time horizon is short. Enterprise AI CX buying decisions are being reshaped this year and next, not over a comfortable decade. Caribbean nations that begin the workforce transition in 2028 will be starting after the enterprise buying decisions have already been made.

Third, the industry framing has an interest in being reassuring. GSAJ's public position that AI is not the primary cause of Jamaica's losses is defensible on the fiscal-year-2026 data, and it may be less defensible on the fiscal-year-2028 data. The productivity, hurricane, and reshoring explanations describe what happened before enterprise AI deployments landed at scale. What happens next is a different question, and the fiscal-year-2026 data does not answer it.

What Caribbean leaders should do

Case 1 · Doomed
Labour-arbitrage BPO is going.

The bottom tier of BPO work is where AI economics are decisive: cost-per-query dropping two orders of magnitude, 2026 enterprise announcements landing at scale, Fortune 500 buyers restructuring at renewal.

What it means for the Caribbean: defending the low-cost seat-count model is designing policy around a market that buyers are already restructuring at contract renewal.

Case 2 · Not Doomed
Complex regulated BPO stays human.

Klarna's reversal, Gartner's 1-in-10 automation projection, and the growing US$358 billion global BPO market all describe an industry that is shifting rather than dying, with humans profitable on the complex tier.

What it means for the Caribbean: the region can win the profitable share of a growing global market by moving up-market.

Case 3 · Both
One version dies, another takes its place.

The version of BPO the Caribbean built is doomed. The version the region needs to build (KPO, AI training, human-in-the-loop) is a growing market.

What it means for the Caribbean: the transition is the whole story, and the transition period is the next five years.

Practical steps · Governments, operators, and workers
The Caribbean's five-year window has already started

For Caribbean governments, the priority is the transition mechanism rather than the peak headline number. That means workforce retraining pipelines linked to CSEC and CAPE exit points, tax incentive structures that reward KPO investment over headcount subsidy, and procurement policy that lets regional providers compete on complex work rather than being routed to the cheapest global bid. The Jamaica BPO recovery task force is the correct instrument. Its execution over the next 24 months is what matters.

For Caribbean BPO operators, the profitable question is what part of the value chain the company owns three years from now. Companies that lock in AI as an internal tool for higher productivity per agent, and that pursue regulated-industry contracts where human accountability is a moat, are positioning for the growing market. Companies whose 2028 pitch to buyers is still price-per-seat are planning for a business that will not exist.

For Caribbean BPO workers, the honest advice is to treat the next 24 months as a re-skilling window. The workers whose careers survive are those who move up the complexity ladder: from Tier 1 support to complex escalation, from customer service to sales and retention, from voice-only work to omnichannel, from agent to AI supervisor. The BPO discipline that made the Caribbean workforce trusted, meaning process rigour, communication skills, and adaptability, is the same discipline that AI operations, data annotation, and human-in-the-loop workflows require.

Frequently asked questions

Directly: mostly no. GSAJ President Yoni Epstein attributes the 2023-to-2026 losses to low productivity, Hurricane Melissa's October 2025 disruption, a tight labour market, and top-tier companies relocating for stronger incentives. That reading is defensible on the data reported so far. Indirectly: the enterprise AI CX deployments that are reshaping global buying decisions arrived in force in mid-2026, meaning the AI displacement thesis has not yet fully tested itself on Caribbean BPO. The next 24 months are the period in which that question gets answered.
It means pure-AI replacement without a strong human escalation path is currently more expensive than the initial headline savings suggested, on the workload types Klarna handles. Both parts of that sentence matter. Klarna proved that AI can absorb roughly two-thirds of consumer-fintech conversations. It also proved that the remaining third is where customer satisfaction is made or lost, and that a pure-AI approach damaged that third badly enough to justify rebuilding the human workforce. Read together: AI absorbs the easy tier at scale, human agents come back for the complex tier, and enterprises have moved from pure-AI replacement to hybrid deployment as their default 2026 strategy.
Grand View Research projects the global BPO market at US$358.6 billion in 2026 and US$695.8 billion by 2033, a roughly 10% compound annual growth rate. The composition inside that growth is shifting toward higher-complexity work in healthcare, financial services, insurance, and regulated industries. Buyers are evaluating providers on technology stack, security posture, and workforce skills, not just hourly rate. The Caribbean's opportunity sits inside that shifting composition.
The most exposed work is high-volume, low-complexity, scripted interactions with structured data: password resets, order tracking, FAQ responses, basic scheduling, and low-tier technical troubleshooting. These are the workloads where AI voice agents already operate at production cost of US$0.05 to US$0.10 per conversation and where enterprise buyers have the clearest business case to switch. The least exposed work is complex, empathetic, regulated interactions with high customer-lifetime-value stakes: insurance disputes, healthcare intake, financial services escalations, retention conversations, and multi-party negotiations.
KPO is knowledge process outsourcing: work that requires domain expertise, judgement, and analysis rather than script following. Examples include legal research, financial analysis, medical coding, market intelligence, insurance underwriting support, and increasingly AI training and evaluation work. KPO pays substantially higher hourly rates than BPO (US$15 to US$50+ per hour versus US$3 to US$6) and is less exposed to AI substitution because the work requires the kind of judgement AI cannot yet reliably provide. The Caribbean's BPO recovery strategies point at KPO because it is the natural next tier for the region's English-speaking, process-disciplined workforce.
The pieces are there. AI training work (data annotation, response ranking, model evaluation, red-teaming) requires English fluency, sustained attention, process discipline, and cultural context, all of which are strengths of the Caribbean BPO workforce. Companies including Annotera are already positioning Jamaica for this work. The barrier is the transition mechanism: retraining the workforce, updating tax and incentive structures, and building the enterprise relationships with global AI providers who buy this work. This is a five-year build rather than a swap of one contract for another.
Probably not in the labour-arbitrage form the peak was built on. Even in the strongest transition scenario, the higher-value work the sector pivots to supports a smaller headcount at higher wages. A realistic 2030 scenario for a successful Jamaica BPO transition is 30,000 to 40,000 jobs across BPO, KPO, and AI operations work, at meaningfully higher average compensation than the 2023 peak. A failed transition scenario is fewer than 20,000 jobs at compressing wages. Which of these Jamaica lands on is decided in the next 24 to 36 months.
Trinidad and Tobago, Barbados, Guyana, and Saint Lucia are earlier on the BPO growth curve than Jamaica. That timing works in their favour because they can design the transition into their sector strategy from the start rather than defending a legacy peak. Practically: prioritise KPO investment attraction over pure headcount incentive, build workforce curricula that treat AI fluency as a baseline requirement, and structure procurement policy so regional providers can compete on complex work. The regional lesson from Jamaica is that the labour-arbitrage form of BPO does not survive intact into an AI economy, and the earlier a country builds around that reality, the softer the transition.
Caribbean BPO has to become the version of itself that AI cannot cheaply replicate: higher-skilled, closer to real KPO, and priced for value rather than hours. The next decade decides whether the region gets there or leaves 40,000 workers stranded on the wrong side of the AI economy.
Caribbean AI Newsletter · The Island AI Brief, 12 August 2026
🌴
About the Caribbean AI Newsletter

The Caribbean AI Newsletter is the leading daily source for artificial intelligence news, analysis, and practical insight for the Caribbean. The newsletter covers policy, workforce, cybersecurity, climate resilience, governance, regional innovation, and the founders building the region's AI sector.

JamaicaTrinidad and TobagoBarbadosGuyanaThe BahamasSaint LuciaGrenadaSaint Vincent and the GrenadinesAntigua and BarbudaDominicaBelizeSurinameHaitiDominican RepublicCuracaoAruba
Visit the Directory
Previous
Previous

Caribbean AI Wrapper Economy: Warning Signs & Checks

Next
Next

Ten Caribbean AI Companies the Region Needs