When IMF And World Bank Treat AI Like A Tsunami, Policy Must Catch Up

The Signal Is The Story

For months, AI labor-market disruption lived in the realm of prediction — speculative essays, fear-driven headlines, cautious corporate planning. The shift that matters happened quietly on a panel in Bangkok.

Kristalina Georgieva, managing director of the International Monetary Fund, described AI hitting the global workforce as a “slow-moving tsunami.” Tharman Shanmugaratnam, Singapore’s president and co-chair of the World Bank’s Jobs Council, said governments must treat employment support not as emergency relief but as permanent structural policy. These are not fringe voices. This is the closest thing the global financial architecture has to a unified signal.

The non-obvious implication: the era of debating whether AI will disrupt jobs is over. The era of deciding who pays for the disruption has just begun.

Who Gets Hit First

Georgieva did not mince words about the earliest casualties. Young graduates entering the labor market face a door that is closing on the bottom rung. The traditional pathway — routine entry-level work, learning on the job, climbing up — relies on tasks that AI can now perform adequately and cheaply.

Shanmugaratnam put a sharper number on it: in the United States, demand for college graduates has fallen for the first time in a century, and the wage premium for a degree is shrinking. That is not a cyclical dip. It is a structural realignment.

The IMF estimates roughly 40 percent of jobs worldwide will eventually be augmented, downgraded, or eliminated. Advanced economies face 60 percent exposure; low-income countries show 23 percent. Do not read comfort into that lower number. AI has simply not arrived there yet. When it does, the exposure curve flattens quickly.

What the numbers obscure is the velocity of change. The tasks AI is automating are not the most dangerous, dirty, or undesirable — they are the ones that once served as on-ramps. Administrative coordination, basic coding, para-professional analysis, customer-facing triage. These are the rungs young workers spent years climbing. Their disappearance creates a hollowed ladder that benefits no one except employers who no longer need to train juniors.

Crisis-Era Means Something Specific

What Shanmugaratnam called “crisis-era” support is not vague good intention. It refers to concrete policy tools deployed during the pandemic and the global financial crisis — wage subsidies, hiring incentives, public-sector bridge employment, active labor-market programs — that are now being repurposed for a different emergency.

The key word in his framing is “permanent.” Pandemic-era measures were always understood as temporary. The argument being made in Bangkok is that AI disruption does not have an end date. It is incremental, task by task, which means the policy response must be too.

This matters because governments have never built institutions for slow-moving labor shocks. They built them for sudden ones. Recession response is familiar territory. A decade-long erosion of entry-level demand is not. That is the institutional gap.

The second-order effect is where this gets uncomfortable. If wage subsidies and hiring incentives become permanent fixtures, they alter the wage floor in ways that ripple through the entire economy. Employers who previously relied on cheap junior labor to cross-subsidize training will need to adjust their cost structures. The social contract around what entry-level work means — as both income source and credential factory — requires rewriting.

The Growth Illusion

Here is where the panel discussion gets uncomfortable. Georgieva acknowledged that AI-driven investment is propping up global growth, potentially adding between 0.2 and 1.0 percentage point to output depending on the region. At the high end, that growth could offset the drag from aging populations in advanced Asia.

Shanmugaratnam pushed back on the easy optimism. Growth and job creation are diverging in China, India, and developing Southeast Asia. The economies are expanding. The jobs are not following.

AI widens that divergence if policymakers stay passive. A productivity boom does not automatically translate into employment — particularly when the productivity gains come from replacing tasks rather than creating new ones.

The divergence matters beyond headline GDP. When productivity rises without corresponding job growth, income concentrates upward. The capital share of income expands. This is not speculative — it is already visible in quarterly earnings reports from firms that cite AI efficiency gains while freezing hiring. The feedback loop is self-reinforcing: cheaper AI deployment leads to lower labor demand, which suppresses wage growth, which depresses consumer demand, which incentivizes further automation to protect margins.

The Institutional Architecture Gap

Neither Georgieva nor Shanmugaratnam addressed this directly, but it is the quiet assumption beneath their remarks: the institutions that govern labor markets — labor ministries, employment services, vocational training systems, university career offices — were designed for a world of stable occupational categories and predictable career ladders. None of that holds anymore.

Career ladders assumed you could enter somewhere, learn, and move up through a sequence of roles. When the first rung vanishes, the entire sequence dissolves. Universities cannot redesign curricula fast enough because the tasks they teach are being automated before the accreditation cycle completes. Employers cannot train on the job when the jobs they were training for no longer exist.

The institutional answer, implicitly, is portable skills infrastructure — systems that track skill development independently of employment, that certify competencies rather than degrees, that provide continuous reskilling rather than one-shot retraining. No country has built this at scale. The Bangladesh and Rwanda models of public-private skills partnerships offer fragments, but the architecture itself remains unbuilt.

Who Wins, Who Loses, What Happens Next

The private sector already generates between 75 and 85 percent of jobs in most countries, according to World Bank President Ajay Banga. His prescription is not state-run employment but de-risking private capital — making investment in hiring young workers less risky for employers. That is a policy architecture that has not been tested at this scale before.

Winners in this scenario: governments that treat AI labor disruption as a structural challenge rather than a future problem. Employers who can design hiring models around augmentation rather than replacement. Workers who can pivot into tasks AI cannot easily replicate.

Losers: a cohort of young people whose first jobs are disappearing before they form. Economies that mistake growth for employment. Policymakers who wait for the tsunami to crest before building seawalls.

The timeline is not distant. Banga estimates 1.2 billion young people will turn 18 over the next 15 years. The private sector must absorb most of them. The question is whether policy catches up before the first wave breaks.

Shanmugaratnam offered one note of cautious hope: AI disrupts task by task, which means there is time — if workers can learn on the job and universities shift toward applied training. That is a condition, not a guarantee. The institutions that enable it — universities, employers, governments — have to choose to change. The signal from Bangkok is that they are finally being asked to.

But the asking is not the doing. The IMF and World Bank have shifted the Overton window on AI labor policy from speculative to structural. What happens next depends on whether governments treat this as a signal to act or merely a signal to monitor. The tsunami is moving slowly. That is the only advantage anyone has left.