Jamaica’s outsourcing sector did not lose 20,000 jobs to artificial intelligence (AI). It lost them to productivity. AI is not what took the work away. AI is how we take it back.

Let me start where the conversation usually ends. Over the last two years Jamaica’s global services sector has gone from roughly 60,000 jobs to roughly 40,000. Earnings have softened. The number of operating companies has thinned. Every version of that story that reaches the public ends in the same place: The robots came.

They didn’t. Not here, not yet, and not in the way the headlines suggest.

The census tells a plainer and more uncomfortable story. We lost work because we became expensive relative to what we produce. Hurricane Melissa cost us continuity. A tight labour market cost us bench depth. Competing jurisdictions out-marketed and out-incentivised us. But underneath all of it sits a single number that we have been reluctant to say out loud: Our output per agent hour has not kept pace with our cost per agent hour. That is a productivity problem. And productivity problems have solutions.

WHAT WE ACTUALLY SELL

We have never sold seats. We have told ourselves we sell seats. Priced ourselves as though we sell seats. And marketed the country as though the value proposition is a chair, a headset, and a time zone. But a buyer in Dallas or Toronto is not purchasing a chair. They are purchasing resolved contacts, collected receivables, processed claims, retained customers.

When a client models Jamaica against Colombia, the Philippines, Honduras, or a domestic AI-first vendor, they are not comparing hourly rates. They are comparing cost per resolved outcome. And on that measure we have been slipping — not because Jamaicans are less capable, but because we have asked Jamaican agents to compete on raw effort while the rest of the world quietly re-tooled.

You cannot win a productivity race with hospitality alone. Our warmth on a call is real and it is a genuine differentiator, but warmth does not compress handle time, and it does not close a $0.15 gap in cost per interaction.

WHERE THE MINUTES ARE GOING

Walk any floor in Montego Bay, Kingston, or Portmore and watch a full shift honestly. A meaningful share of every hour goes to work that produces nothing the client is paying for:

* After-call work: Agents typing summaries, tagging dispositions, updating CRM fields by hand.

* Knowledge search: Agents hunting through outdated intranets and PDFs mid-call while a customer waits.

* Repetitive tier-one volume: Balance checks, password resets, order status — high volume, low value, and yet staffed with trained people who could be doing complex work.

* Sampled quality: Supervisors reviewing two or three calls per agent per month and calling that a quality programme.

* Forecast error: Interval-level over- and under-staffing that shows up as abandonment on one side and idle payroll on the other.

None of that is a Jamaican failing. It is a global legacy of how contact centres were built. But it is precisely the layer that modern AI removes — and every hour we leave in it is an hour we are paying full price for and selling at a discount.

THE MEASURABLE CASE

This is no longer speculative. The efficiency gains are documented in production environments, not vendor decks:

* Real-time agent assist — next-best-action prompting, live intent detection, surfaced knowledge — has been shown to reduce average handle time by roughly a quarter (Metrigy research, cited via Genesys).

* Automated after-call work, where generative AI writes the interaction summary and populates the CRM, cuts wrap time by around a third (Metrigy, via Zoom).

* AI-driven quality assurance moves review coverage from a 2 per cent sample to effectively 100 per cent of interactions, which changes coaching from anecdote to evidence.

* AI-assisted workforce management tightens interval-level forecast accuracy, which is where most of our unrecovered payroll actually leaks.

* National Bureau of Economic Research work on customer service teams using AI support found average productivity gains in the mid-teens — and, critically, the largest gains accrued to newer and lower-tenured agents.

Hold that last finding up against Jamaica’s situation. Our binding constraint is a thin, young talent pool in a low-unemployment economy, with long ramp times and attrition that eats training investment before it pays back. The single best-evidenced effect of AI in this industry is that it compresses the distance between a new hire and a competent one. That is not a threat to Jamaican employment. That is the most direct answer to our labour constraint that anyone has put on the table.

THE HONEST CAVEAT

I am not selling a miracle, and this industry has heard enough of those. AI will not fix a broken operation. It will make a good process faster and a bad process visibly broken, which for those of us who run these businesses is uncomfortable but ultimately useful.

Deployments fail routinely, and they fail for predictable reasons: Bolt-on tools that cannot reach real-time data, no source-grounding, no governance, no control group, and no honest measurement of whether time saved in summaries is simply reappearing as correction work later.

So the standard has to be discipline, not enthusiasm. Start with the unglamorous use cases — post-call summarisation, knowledge retrieval, QA coverage — where the human stays in the loop and the risk is low. Measure against a control group. Demand production deployment data from vendors, not demos. Track containment, first-contact resolution, handle time, and CSAT together, because a system that optimises one while quietly destroying another is worse than no system at all.

WHAT THIS ASKS OF US

If AI is the productivity fix, then the work is not primarily technological. It is institutional, and it splits three ways.

Operators must stop treating AI as a cost-reduction lever and start treating it as a margin and mix lever. The point is not to run the same work with fewer Jamaicans. The point is to make each Jamaican hour worth more, so we can bid credibly for the complex, regulated, judgement-heavy work — the health-care, financial services, and technical support programmes that pay two and three times what basic voice pays, and that nobody offshores to a destination they think of as cheap.

Government must recognise that our incentive and marketing regime is competing against jurisdictions that never stopped selling. We have world-class fundamentals — proximity, culture, accent neutrality, English-language depth, a genuine service instinct — and we have been under-marketing them for years while our competitors bought the mindshare.

Alongside that, the enabling infrastructure has to be treated as national economic infrastructure — reliable power, resilient telecoms, and post-disaster continuity, because Hurricane Melissa proved that our exposure on that front is priced into every deal we bid.

The talent pipeline must be rebuilt around what the job is becoming rather than what it was. If AI handles tier one, then the entry-level job is no longer script-reading — it is exception handling, judgement and empathy, supervising AI output rather than competing with it. That is a different curriculum, and HEART/NSTA Trust, the tertiary institutions and industry need to be designing it together now, not in three years.

There is a version of the next five years in which Jamaica sits still, keeps insisting the problem was AI, and watches the number go from 40,000 to 30,000 while the work migrates to destinations that re-tooled. And there is a version in which we are honest that the problem was productivity — and then use the most powerful productivity technology ever built to close the gap, move up the value chain, and rebuild toward 60,000 jobs that are better paid and harder to displace than the ones we lost.

The window for that choice is open. It is not open indefinitely. Every quarter we spend arguing about whether AI is coming is a quarter our competitors spend deploying it. We did not lose this on talent. We lost it on output. And output is the one thing we can fix.

Yoni Epstein is president of Global Services Association of Jamaica as well as founder & chairman of itel