1.89 million BPO workers, 53,500 training seats: The Philippines’s AI maths doesn’t add up

Industry bodies rarely talk down their own industries. So when the IT and Business Process Association of the Philippines (IBPAP) rewrote its 2028 roadmap in July, the revision said more than any press release could. The workforce target fell from 2.5 million to a best-case 2.14 million, and the revenue goal dropped from US$59 billion to US$50.5 billion.

IBPAP blamed two things: AI productivity tools on the client side and fiercer global competition.

Also Read: The Philippines never lacked talent, it lacked leverage and AI is changing that

That is the backdrop to Gobi Partners’s report Solo Horizons: Redefining Scale for the Philippine AI Economy, which names IT-BPM as the country’s largest single-sector disruption exposure. The sector is a big thing to put at risk. In 2025, 1.89 million workers generated US$40.3 billion, and much of the country’s new urban middle class was built on headset shifts timed to American mornings.

The first part of this series looked at how AI-native one-person companies might stand in for a vanished Series A. This instalment turns to the people expected to found them. It also looks at the gap between the founder pitch and what retraining can actually deliver.

The jobs most exposed are the ones the industry was built on

Estimates of the damage vary widely, which is telling in itself. Consultancy Kearney once flagged up to 100,000 BPO jobs at risk. Labour-market analyses by Swarm and tech strategist Dominic Ligot put about 680,000 Filipino workers in high-risk occupations, or 1.6 per cent of the national workforce. Of those, 306,000 face outright displacement by 2030.

The breakdown reads like a BPO org chart. About 245,000 call-centre and voice CX agents face a 67 per cent automation probability as voice agents take over routing and first-line resolution. Some 89,000 data-entry and back-office clerks sit at 78 per cent, and 45,000 bank and insurance processing clerks at 72 per cent. The list also counts 156,000 cashiers and retail processors, who are not BPO workers at all. That is a reminder that headline figures blend very different labour markets.

International Monetary Fund microdata suggests nearly a third of Filipino workers are exposed to AI in some form, with BPO holding the highest concentration of jobs at risk of replacement. The report places the peak disruption window in 2026 and 2027. In other words, now.

A retraining system running six years behind

This is where the report is bleakest. The Philippines needs to reskill around 150,000 workers a year. Government agencies, universities and corporate programmes together can train about 53,500. That leaves 96,500 people a year without a seat. By Gobi’s estimate, it puts the pipeline roughly six years behind the disruption peak.

Employers are not closing the gap. Among Philippine enterprises adopting AI, the median annual training budget is US$75,000, and only 34 per cent of employees say they have had formal AI instruction. The problem runs deeper than corporate thrift: the AI divide in the Philippines began long before AI, in classrooms where basic literacy still lags.

Even the optimistic arithmetic has losers. The report projects 450,000 new AI roles nationwide by 2030, a theoretical net gain of 144,000 jobs. But Metro Manila is forecast to gain a net 67,000 high-value roles, while regional BPO hubs lose 89,000 as offshore clients consolidate.

Also Read: India is showing the world what AI will do to jobs and education

The divide also runs along gender and pay lines. Women make up 58 per cent of the BPO workforce, while the emerging machine-learning and data-ops roles skew 68 per cent male. A displaced BPO worker earns about PHP270,000 (US$4,700) a year. Specialist AI engineers command PHP1.8 million to PHP3.2 million (roughly US$31,000 to US$56,000), the kind of premium that has made agentic AI Singapore’s hottest job skill too. A net gain is cold comfort to a woman on a provincial voice account who falls on the wrong side of all three lines.

Why Filipino agents are better placed than most to jump

Set the gloom aside and the workforce has real advantages. The Philippines topped the 2026 Ataraxis Global Outsourcing Talent Index across 193 countries with a score of 90.65, ahead of India and Malaysia. It won largely on English proficiency, scoring 90 to India’s 60, and on cultural fluency with Western clients. The same strengths explain why Filipino virtual assistants dominate the global remote workforce.

Workers are also not waiting for permission. Bain & Company and Google data show Gemini prompts from the Philippines for customer-support content run at nearly three times the Southeast Asian average. That fits a pattern of Filipinos using AI as a burnout shield. Microsoft’s Work Trend Index finds 83 per cent of Filipino AI users bring their own tools to work, ahead of any company mandate.

That is the basis of Gobi’s “1% Founder Thesis.” If one per cent of the 1.89 million IT-BPM workers became founders, the country would gain roughly 18,900 AI-native ventures, more than the 14,000 startups Tracxn tracks nationally. A team lead who has spent a decade inside a US insurer’s claims workflow knows things no model vendor does.

But a worker who brings Gemini to her desk to survive a quota is not automatically a founder. The thesis also depends on the very training seats the system cannot supply.

TESDA’s AI budget is smaller than the problem

Gobi’s prescriptions start with public skilling. TESDA has a record PHP19 billion (about US$330 million) budget for 2026, of which PHP500 million (about US$8.7 million) is set aside for AI micro-credentials. That is 2.6 per cent of the budget, going to the skill the whole sector now depends on. The TESDA-Microsoft partnership targeting 100,000 women and DepEd’s one-million-learner AI literacy drive at least aim at the gender and regional gaps.

Second, the report says outsourcing firms must move from headcount billing to outcome-based, AI-augmented delivery. Early adopters report 28 per cent efficiency gains, and 78 per cent see a positive return from AI copilots that assist agents rather than replace them. The catch is that a client getting 28 per cent more output per seat will eventually ask why it needs every seat.

Also Read: The Philippines does not need to build AI to have an AI advantage

Third, investors should back “agency-to-SaaS” operators. These are BPO alumni building workflow tools, automated support systems and Taglish language models such as Senti AI’s Balagtas, part of a new crop of Philippine AI startups. It is a sensible thesis in a market where, as part one showed, Series A money has nearly disappeared.

The report is right that the disruption window is already open. It is less convincing that a one per cent founder conversion can offset a training shortfall of 96,500 people a year, especially while neighbours race ahead; Malaysia wants 300,000 AI jobs by 2030. Turning call-centre veterans into builders is a worthy goal. The harder and more urgent job is making sure the other 99 per cent are not written out of the Philippines’ next economy.

The post 1.89 million BPO workers, 53,500 training seats: The Philippines’s AI maths doesn’t add up appeared first on e27.