The Philippines Has 12.7 Million AI-Exposed Jobs. But Exposure Is Not Job Loss.

The useful question is no longer whether AI can affect Philippine work. It can. The harder question is whether exposed jobs are shrinking, growing, or changing, and which workers are seeing the effects first.

Start with what exposure actually means

The International Labour Organization estimates that more than one-quarter of employment in the Philippines, or 12.7 million jobs, has more than minimal potential exposure to generative AI. A July 2026 ASEAN update puts the Philippine share at 28.1 percent, the second highest among the nine ASEAN countries with comparable data after Singapore.

That does not mean 12.7 million jobs are expected to disappear. The ILO explicitly distinguishes occupational exposure from job replacement. Only 3.6 percent of Philippine jobs fall into its highest exposure category, where the risk of displacement is elevated. For most exposed occupations, the more plausible near-term effect is that some tasks are automated or augmented while the job itself changes.

This distinction matters because exposure is a property of the tasks inside an occupation. Employment is an observed labor-market outcome. They are related, but they are not the same variable.

The Philippine labor market is already giving mixed signals

The Philippine Statistics Authority estimated 50.66 million employed people in June 2026. Services accounted for 62.7 percent of employment. Unemployment stood at 4.9 percent and underemployment at 12.1 percent.

The sector movements are more interesting than the national totals. Administrative and support service activities added an estimated 456,000 jobs year on year in June, one of the largest increases in the economy. Financial and insurance activities moved in the opposite direction, losing an estimated 195,000 jobs year on year.

Those observations should not be interpreted as evidence that AI created one change and caused the other. The Labor Force Survey is descriptive, not a causal study of AI. But the contrast shows why a simple exposure ranking is insufficient. A highly exposed sector can continue to add workers. Another can shrink. Demand, demographics, business cycles, offshoring, wages, regulation, technology and firm behavior can all matter at the same time.

Clerical work is the clearest Philippine watchpoint

The ILO estimates that 93.7 percent of Philippine clerical roles are exposed to GenAI, with 37.8 percent in the highest exposure group. These occupations include data entry, record keeping and administrative support, where work contains many structured information tasks that current AI systems can perform or accelerate.

The exposure is also concentrated by industry. Nearly nine in ten Philippine finance and insurance jobs are estimated to have some GenAI exposure. Women face roughly twice the overall exposure rate of men because they are more concentrated in clerical, administrative and service roles. In the National Capital Region, around two in five jobs are exposed.

These are useful risk indicators. They are not forecasts of layoffs.

What forecasters expect in the United States

Metaculus runs a Labor Automation Forecasting Hub that aggregates probabilistic forecasts about the US workforce. As of 6 September 2026, its community forecast expects the most AI-vulnerable occupations in its sample to shrink 12.9 percent by 2035 while the least vulnerable group grows 7.1 percent.

Among its representative US occupations, the 2035 forecasts include financial specialists at -14.2 percent employment versus 2025, software developers at -12.2 percent, services sales representatives at -10.1 percent, construction workers at +4.3 percent and registered nurses at +11.4 percent.

These are US forecasts. They are not Philippine projections. Labor demand, wage levels, outsourcing, demographics, occupational structure and AI adoption differ too much for us to copy those percentages onto Philippine workers.

Their value is as an external forecasting benchmark. They tell us what informed forecasters currently expect in comparable categories, and they give us hypotheses that Philippine data can confirm, reject or complicate over time.

A better Philippine AI jobs dashboard

We think a useful Philippine monitor should keep three evidence layers separate.

Observed outcomes. Employment, unemployment, underemployment, hours, wages and occupational or industry shifts from official Philippine sources such as the PSA.

Exposure. Estimates of which occupations contain tasks that GenAI can automate or augment, using research such as the ILO exposure index.

Forecasts. Explicit probability distributions about future employment and labor-market outcomes. Metaculus provides a useful external benchmark today. Philippine-specific forecasting questions would be stronger.

The classification layer matters too. The Philippine Standard Occupational Classification is patterned after ISCO-08, which gives us a reproducible basis for mapping Philippine occupations to international evidence. Where US occupational forecasts are used for comparison, the mapping should be explicit and its quality reported rather than hidden.

What we will watch next

The strongest warning signal would not be high AI exposure by itself. It would be several indicators moving together: persistent employment decline in exposed occupations, weaker entry-level hiring, reduced hours, falling real wages, rising underemployment, or a shift toward smaller senior teams with fewer junior pathways.

The opposite is also possible. Some exposed occupations may grow because AI raises worker productivity, lowers service costs or expands demand. That is why the monitor has to follow outcomes rather than assume them.

The PSA is scheduled to release the July 2026 Labor Force Survey on 8 September 2026. That gives us the next observation almost immediately.

Methodology note

Pukot Research treats observed Philippine labor-market data, AI exposure estimates and forecasts as separate evidence classes. We do not infer that AI caused a change in Philippine employment from contemporaneous Labor Force Survey data alone. We do not apply US Metaculus occupation forecasts directly to Philippine occupations. Cross-market comparisons should use documented occupational crosswalks, ideally US SOC to ISCO-08 to PSOC, and should report mapping uncertainty. ILO exposure measures describe potential task exposure, not a predicted probability that a worker will lose a job.

Sources

Philippine Statistics Authority, June 2026 Labor Force Survey
International Labour Organization, Generative AI and jobs in the Philippines
ILO, Navigating Generative AI transformations in ASEAN labour markets
Philippine Standard Occupational Classification
Metaculus Labor Automation Forecasting Hub