IMF chief warns of a "slow-moving tsunami" from AI for labor markets and tax revenue

International Monetary Fund Managing Director Kristalina Georgieva has warned that artificial intelligence threatens labor markets and tax systems. She urged governments to act now rather than after the first mass disruptions. The image she chose is a "slow-moving tsunami": the wave is visible well in advance, but the closer it gets to shore, the less time remains to prepare. For context, in January 2024 IMF analysts estimated that AI would affect about 40% of jobs worldwide and roughly 60% in advanced economies.

One caveat: the published announcement of the speech contains no new calculations or specific measures. It is a political signal, not a new report. But the signal comes from an institution whose recommendations finance ministers read closely.

Why "slow" is still a tsunami

The word "slow" is doing real work here. AI does not wipe out professions in a single quarter. It takes over tasks gradually: first draft analysis, customer support, routine code, translation, document preparation. Then companies revisit headcount and hiring requirements for newcomers. The most vulnerable segment is entry-level positions, because that is where people used to be trained on routine work.

In earlier work, the IMF stressed that AI's impact is uneven. Economies with a high share of office and knowledge work are hit hardest. Emerging markets are less exposed, but they also have fewer opportunities to take advantage of the technology. The gap between countries may widen.

Taxes: budgets are tied to wages

The second part of the warning gets less attention than the fear of unemployment, but it is more acute for governments. In advanced economies, a significant share of budget revenue, about half according to the OECD, comes from income tax and social contributions. All of it is tied to paid labor.

If some tasks move from people to models and servers, the tax base shrinks. Profits may still grow, but they accrue to capital owners and technology platforms, which are taxed differently and are often mobile across jurisdictions. The arithmetic is simple: spending on benefits, retraining and pensions rises, while revenue from labor fails to keep pace.

It is not yet clear which tools governments will choose. Three options are usually on the table:

• shifting the tax burden from labor to capital;
• revising taxation of corporate profits and digital services;
• taxes on consumption and resources, including data center energy use.

Each option carries a price in the form of competition for investment and political risk.

Education as the first lever

Georgieva named education as the area that needs rethinking first. The logic is clear: schools and universities operate on a 10–15 year cycle, while model capabilities change in months. Programs built around memorization and routine skills risk preparing people for tasks that are already automated.

Key figures and facts from the article:

• 40% of jobs worldwide are exposed to AI (IMF estimate, 2024);
• about 60% of jobs in advanced economies;
• two risk areas: employment and tax revenue;
• the main appeal: rethink education and begin adapting immediately.

The announcement gives no details on new forecasts or specific instruments.

What to do in practice

For companies, a sensible step is to assess which tasks in each role will be automated first and to build retraining paths in advance, while that is still cheaper than mass layoffs. Junior positions deserve particular attention: if interns and trainees disappear, in a few years companies will have to figure out where to find experienced specialists.

For professionals, the practical defense is to move from performing routine operations to defining tasks, checking results and taking responsibility for them. For now, these skills remain with humans.

For investors and CFOs, the tax trend means regulatory risk: the faster AI erodes taxes on labor, the more likely a tougher regime for capital and data centers becomes. It is better to build such scenarios into models now.