# The AI pipeline problem: How automation is breaking the career ladder in Southeast Asia <div class="pills-container"> <span class="pill">Published: May 6, 2026</span> <span class="pill">Reading Time: 5 minutes</span> <span class="pill">This is a linkpost to <a href="https://e27.co/the-ai-pipeline-problem-how-automation-is-breaking-the-career-ladder-in-southeast-asia-20260504/">e27</a></span>. </div> Rishabh Mishra enrolled in one of India’s top engineering colleges in 2022 with a plan that felt safe to study computer science, write code, and then get a job after graduation. Three years later, fewer than a quarter of his 400 classmates had secured job offers. “Everyone is so panicked, even our juniors,” [he told Rest of World.](https://restofworld.org/2025/ai-job-losses-engineering-graduates/) The same pressure is showing up globally. Entry-level hiring at big tech companies has fallen by more than [50 per cent over the past three years](https://restofworld.org/2025/ai-job-losses-engineering-graduates/). In Singapore, one of the most AI-ready economies in SEA, the full-time permanent employment rate for graduates sat at 51.9 per cent in 2025. SMU’s own provost [acknowledged](https://www.straitstimes.com/business/more-fresh-grads-found-jobs-in-june-2025-than-a-year-ago-mom) that “hiring conditions are becoming more cautious as organisations manage uncertainty and accelerate automation.” Most of the conversation about AI and automation frames it as a productivity story. AI would handle routine work so humans could move up to higher-value tasks. What that framing misses is that entry-level roles are also where people develop the judgment needed for mid- and senior-level work. Automating them creates a pipeline problem that only becomes visible years later. ## AI causes a job pipeline problem that nobody is talking about The Philippines has nearly two million workers in Business Process Outsourcing (BPO), a sector worth [US$35 billion and making up eight per cent](https://www.straitstimes.com/asia/asean-is-already-economically-divided-ai-will-make-it-worse) of the Philippine GDP. For many of those workers, the BPO sector offers the first step to a potential career. However, [BPO jobs are also the most automatable](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025043-print-pdf.pdf). A worker two years into a customer service role has learned things that are not in any training manual, like how a client escalation actually gets resolved, when to push back on an unclear instruction, or how to make a judgment call when the process does not cover the situation. These are skills built through repetition on low-stakes tasks, where the cost of a mistake is recoverable. However, this path to skills development is now under direct pressure due to AI and automation. The International Labour Organisation (ILO) published an [analysis](https://www.ilo.org/resource/employment-focus-blog/navigating-generative-ais-transformations-asean-labour-markets) in April 2026 which found that 93.7 per cent of clerical roles in the Philippines are already exposed to generative AI (GenAI), with 37.8 per cent of which being in the highest-exposure category and facing direct automation risk. Tasks like data entry, record-keeping, and structured administrative work, which are exactly what junior workers typically spend their first two years doing, are categorised as high-exposure. A peer-reviewed [study](https://doi.org/10.46557/001c.132415) in Asian Economics Letters examining ASEAN labour markets from 2015 to 2020 found that higher AI exposure already correlated with declining employment shares and wages across the region. The Philippines was among several countries where researchers identified a “reinstatement effect.” That means that AI will eliminate certain tasks but will also create new ones within the same roles. For example, a customer service agent who used to spend half their shift logging call summaries now supervises the AI doing it instead. The optimistic read is that [AI will create roughly 170 million new jobs this decade](https://www.weforum.org/stories/2025/04/ai-jobs-international-workers-day/). But the jobs most at risk are entry-level, and the jobs being created tend to require experience that entry-level work is supposed to build. This is essentially a chicken-and-egg problem.  The World Economic Forum data shows [AI could replace more than 50 per cent of tasks](https://www.weforum.org/stories/2025/04/ai-jobs-international-workers-day/) performed by market research analysts and sales representatives, compared to just nine per cent and 21 per cent for their managerial counterparts. The [ILO’s 2026 Employment and Social Trends report](https://www.ilo.org/publications/major-reports/world-employment-and-social-outlook-trends-2026) noted that global labour markets look stable at the headline level while inequalities widen underneath, with AI disruption falling hardest on young workers entering structured, task-based roles, like the ones in the BPO sector. The common counterargument is that new tasks will appear inside these roles, but that is not the same as the learning paths within these careers staying intact. If the tasks being automated are the ones that teach judgment, and the tasks left behind already require it, then junior workers end up being handed work they are not yet equipped to do. There may still be jobs, but fewer of them will actually develop the people doing them. That will likely lead to a supply of jobs that not enough people are qualified to do. ## Most of SEA lacks the institutional capacity to manage this transition Singapore has roughly 3.5 AI professionals per 1,000 workers. The Philippines, Thailand, Indonesia, and Vietnam each hover around 0.2 professionals per 1,000 workers. When entry-level roles change, someone inside the organisation needs to figure out what to replace them with and how to keep junior workers developing. Singapore has enough people who can do that. Most of the rest of SEA does not. This means the cost of getting automation wrong falls harder outside Singapore. The savings from automating entry-level tasks show up immediately, but the problems it will cause will show up five to seven years later. It will show up when the workers who should have grown into mid-level roles turn out to be less experienced than expected, because the jobs that used to build that experience no longer exist. Access Partnership estimated that 57 per cent of SEA’s workforce, or around 164 million people, could see their jobs reshaped or disrupted. ILO notes that women and young workers are concentrated in clerical and administrative roles, so they face the most AI exposure. In Indonesia, generative AI exposure among workers aged 15 to 24 runs at 26.1 per cent, compared to 21.1 per cent for older workers. These are the workers who most need entry-level jobs to build toward something, and they are the ones most likely to find those jobs gone. ## The question worth asking before you automate Cutting headcount is not the same as having a plan. The better question to ask if you’re an organisation planning to automate workflows is to think through what junior employees are actually learning right now, and will that be enough to produce the experienced mid-level workers you need in five years? The Philippine IT-BPM sector has pointed toward training workers to supervise AI tools, manage automation workflows, and support decisions rather than just execute transactions. That instinct is right, but announcing a retraining program and actually building a career path that develops people are two different things. The second one requires understanding what entry-level jobs were actually teaching in the first place. That is the part most organisations will skip, and the ones that do will find out its consequences five years too late.