Eighty-nine percent of Philippine BPO workers are classified as highly exposed to AI. That figure has appeared in policy briefings, job-fair warnings, and newspaper headlines since IMF Working Paper WP/2025/043 published it in February 2025. What rarely travels with it is the number that belongs right next to it: 61% of those same exposed roles are rated highly complementary to AI, meaning the technology is more likely to assist the worker than displace them. Both figures come from the same paper. The first got picked up widely. The second mostly did not.

This analysis works through what the available data says, role by role, as of August 2026.

What IMF Working Paper WP/2025/043 Actually Found

The paper, by Micholo Cucio and Tristan Hennig, combined Philippine labor force survey microdata with occupational-level measures of AI exposure and AI complementarity. The headline result: around one third of all Filipino workers are highly exposed to AI, with BPO at the far end of the distribution. Eighty-nine percent of BPO workers are classified as highly exposed.

Exposure, in this methodology, measures whether AI can perform a meaningful share of the tasks in an occupation. It does not measure whether an employer will cut the job. That is where complementarity enters the analysis. Sixty-one percent of those highly exposed BPO roles are also rated highly complementary, meaning the AI is expected to raise the value of the human worker rather than substitute for them. A role that is both highly exposed and highly complementary is one where learning to use AI well is the path to greater productivity, not to redundancy.

The exposure-versus-displacement distinction matters for anyone who has seen briefings built on the 89% figure alone. Exposure tells you where AI touches the work. Displacement tells you whether the job disappears. They are correlated but not the same outcome, and the 61% complementarity rate is precisely why the two cannot be used interchangeably.

Displacement Is Narrower Than Transformation

Coverage of the IMF findings in the Daily Tribune put the displacement-specific estimate at about 14% of Philippine jobs facing high displacement risk. The broader group, 22% of jobs expected to undergo significant transformation, is substantially larger. These are different planning problems. Transformation requires retraining within an existing employment relationship. Displacement requires new job creation and career-transition infrastructure. Treating both groups with the same response wastes resources on the first and underprovides for the second.

The ILO's February 2026 country brief on the Philippines, focused on generative AI specifically, found that more than a quarter of Philippine employment, roughly 12.7 million workers, is exposed to generative AI, the highest rate among comparable ASEAN economies at the time. But the same brief found that only 3.6% of total employment falls in the highest exposure category where displacement risk is most concentrated. As reported by BusinessMirror citing the ILO brief, the expected impact will arrive through task transformation within occupations rather than through mass job replacement.

The Philippines in the Regional Picture

The ILO's July 2026 regional follow-up placed the Philippines second in ASEAN for generative AI exposure, with 28.1% of employment in occupations with more than minimal GenAI exposure, behind only Singapore at 42.2%. The same brief noted that widespread disruption was not yet visible and that employment in highly exposed occupations was still growing across the region.

The number in that brief that deserves more attention from national agencies is not the exposure rank. It is the preparedness rank: the Philippines is second in ASEAN for AI exposure but fourth in preparedness. A country that leads on exposure and lags on readiness faces a wider adjustment gap than either number alone conveys. That gap, not the exposure percentage, is the structural planning challenge for DICT, TESDA, and DOLE.

Employment Still Growing, but Entry-Level Hiring Is Shifting

After the IMF paper appeared in February 2025, the contact center industry reported it had added more than 60,000 workers (self-attested by the industry as of mid-2026). Aggregate headcount was still rising roughly eighteen months later. This is consistent with the IMF and ILO findings, not in tension with them. Transformation does not require immediate net job loss; it changes which tasks a worker spends time on and which skills command a hiring premium.

What aggregate employment data does not surface is what Santiago and Company identify as quiet decoupling: employment still growing, but entry-level hiring weakening and revenue growth increasingly separating from headcount growth. If that pattern holds, it is the early signal that adjustment is already underway at the hiring margin even while headline totals appear stable.

What the Job-Board Snapshot Shows

BPOAI.ai ran a posting-count snapshot using site-restricted Google searches across JobStreet PH, Indeed PH, and LinkedIn Jobs for the weeks of 9 August and 16 August 2026. The method counts distinct indexed posting URLs per search term and approximates active indexed listings, not board-side totals. Results should be read as a directional signal, not a comprehensive count (BPO AI first-party analysis of JobStreet, Indeed, and LinkedIn data, August 2026; confidence 0.70, inferred/modelled).

Within this sample, AI-adjacent titles outnumber traditional voice-CSR titles roughly five to one on the one platform where either type appears at all. That directional signal fits the IMF complementarity framing: the roles absorbing AI-adjacent work are advertising; the roles most exposed to task automation are not posting at the same rate. For a broader view of active employers and their current service lines, the BPO directory tracks company-level footprint across the sector.

What the Available Data Supports by Role

The question most often asked about this sector is whether AI will replace call center jobs in the Philippines. The data supports a more precise answer than the headline percentages have offered:

  • Highest displacement risk: Scripted, high-volume voice roles where AI handles most task content without requiring human judgment at handoff. The IMF exposure methodology places these in the displacement tail, and job-board data shows near-zero indexed demand for traditional voice titles on the platforms surveyed.
  • High exposure, high complementarity: Quality assurance, complex escalation handling, account management, and judgment-intensive back-office roles. The IMF rates 61% of all exposed BPO roles in this category. The most likely outcome for this group is task reallocation rather than job elimination.
  • Emerging demand: AI trainer, AI quality analyst, prompt engineer, conversation designer, AI operations specialist. All five showed consistent LinkedIn activity across both snapshot weeks while traditional voice titles indexed near zero on the same platform.

A gap worth naming openly: the SSRN paper by Nartey and the IBPAP role-level workforce projections cited in the background brief for this article could not be verified from the research material available at publication. The IBPAP estimates of up to 36% of positions at risk and approximately 53,000 roles affected are held pending retrieval of the primary documents. The IMF, ILO, and job-board data above establish the directional picture; they do not support precise occupation-level displacement percentages without those papers.

The institutional pacing problem sits behind all of this. Santiago and Company estimate that AI reached measurable economy-wide business adoption in roughly 14 to 15 months, while the median national workforce reform takes 59 months to reach its first operational milestone. That is approximately a fourfold gap between the deployment cycle and the institutional response. The exposure data is available now. The 59-month clock is already running.