Fifty-Nine to Forty-Three

The Philippine IT and business process management sector is the largest employer of English-speaking service labour outside the United States and India, and worth roughly 8 percent of Philippine GDP. In 2022 its trade association published a six-year roadmap: 59 billion dollars of revenue and 2.5 million jobs by 2028.

The midterm revision arrived in July. Revenue by 2028: 43.3 to 50.5 billion dollars. Employment: 1.85 to 2.14 million. The sector employs about 1.9 million people now, which means the floor of its own two-year forecast sits below its present headcount.

It was widely reported as artificial intelligence arriving in the world's call centre capital. The association's own arithmetic does not say that, and the gap between the two readings is the most instructive thing about the episode.

Divide before concluding

An automation shock has a measurable signature: output holds or grows while headcount falls, so revenue per worker rises. That is what displacement looks like in a services account.

It is not what was published. The 2022 plan implied roughly 23,600 dollars of revenue per employee. The revised pessimistic case implies about 23,400. The revised optimistic case implies about 23,600. Three scenarios drawn four years apart, spanning a 16 billion dollar revenue range, and all of them within about one percent of the same figure for output per worker.

Both lines were scaled down together. The association did not forecast doing the same work with fewer people. It forecast less work, at an unchanged assumption about what a worker produces.

The near term runs the same way. 2025 closed at just over 40 billion dollars on 1.9 million workers, about 21,000 each. 2026 is projected at 42.3 billion on 1.96 million, about 21,600. Growth in output per worker of roughly 2.5 percent, which in a services industry is an ordinary year.

The association said this and was not heard

Jack Madrid, IBPAP's president and chief executive, declined the automation framing when the revision was published. "I think AI is real, but we haven't seen it scale yet," he said. "It affected some jobs, but redeployed employees have transitioned to other roles." He attributed the downgrade to macroeconomics and geopolitics, with client firms taking longer to decide where to place work.

There is an obvious incentive reading, which is that a trade association would prefer not to announce its members are automating away their own labour force. It is worth stating and then setting against the arithmetic, which is not a matter of framing. Had the association wanted to disguise displacement it would have had to hold the revenue line while cutting headcount, and it did neither.

The simplest account consistent with both the quotes and the numbers is the dull one. Demand for offshore service capacity is being committed more slowly, in a period of tariff uncertainty and reshoring pressure, and a forecast written before any of that has been marked to reality.

Why the misreading matters for policy

Displacement and deferral produce similar-looking labour markets two years out and call for entirely different instruments.

Displacement is an event. It has a date, an employer, an identifiable set of workers, and it triggers the machinery built for industrial transition: notice periods, severance, retraining entitlements, adjustment assistance. Every one of those instruments is keyed to something a firm does to someone.

Deferral is the absence of an event. Nobody is dismissed. The positions are simply never opened, which produces no notification, no dispute, no administrative record, and no trigger for any transition programme in existence. It shows up years later in the age profile of an industry and in the share of a graduating cohort that never entered it.

A government that reads this revision as automation will fund retraining for displaced BPO workers who have not been displaced, and will miss the cohort that never got hired. That is not a hypothetical failure mode; it is the standard one. Trade adjustment programmes have historically reached a small fraction of the workers they were designed for, largely because eligibility is defined by a separation event.

The counter-case

India, the other large offshore services economy, is moving the opposite way. Naukri's JobSpeak series records fresher hiring, the zero to three year band, up 17 percent year on year in February 2026 and 16 percent in March, with the IT sector up 6 percent overall and AI and machine learning roles up 45 percent across the fiscal year.

Same technology, freely available in both markets, opposite entry-level trajectories. That is a strong argument against any account in which the technology alone determines the employment path, and a reason to treat confident projections in either direction with suspicion.

What is left when the framing is stripped out

Two things, and they are worth more than the headline was.

The first is that an industry association with every reason to project confidence has published, in a document intended to attract investment, a two-year employment range whose lower bound is below its current workforce. That is a real signal about expected demand for labour-intensive offshore services, and it does not depend on why.

The second is a caution about the season. A great many organisations are about to attribute ordinary demand weakness to artificial intelligence, because that framing is available, fashionable, and flattering to management. The arithmetic that separates the two is division, it takes one line of a spreadsheet, and in this case the most-cited AI-and-jobs datapoint of the summer does not survive it.

Sources

• Philippine Daily Inquirer, "Philippine BPOs slash growth goals on AI shift", 15 July 2026. Source for the 2022 roadmap of 59 billion dollars and 2.5 million jobs by 2028, the revised 2028 ranges of 43.3 to 50.5 billion dollars and 1.85 to 2.14 million jobs, the 2026 projection of 42.3 billion dollars and 1.96 million workers, the 2025 actuals, and both quotations from Jack Madrid, IBPAP president and chief executive. Every revenue-per-employee figure here is this publication's arithmetic on those published numbers. The association did not present the ratio and does not characterise its revision as we have.
• Naukri JobSpeak, February 2026 and March 2026. Source for Indian fresher hiring up 17 and 16 percent year on year, the IT sector up 6 percent, and AI and machine learning roles up 45 percent across the fiscal year.
• The Economist, issue of 8 to 14 August 2026, for raising the Philippine industry's response to AI. Its figures for the Indian IT market, demand down 3 percent with AI roles up 25 percent, do not reconcile with Naukri's published series; we have used Naukri.