The IMF finally names the winners, the losers, and the clock
On the eve of Davos, Kristalina Georgieva pressed publish and, with a neat blend of timing and data, shifted the AI-and-work debate from vibes to measurements. The IMF didn’t opine about the future; it instrumented the present. In a blog post paired with a 49-page Staff Discussion Note, the Fund traced how AI-era skills have already rewired hiring, wages, and headcounts. The headline message is deceptively simple: prepare people to “use AI rather than compete with it.” The second message is the one that matters for everyone who reads this newsletter: whether that advice turns into jobs or job loss depends on which skills spread, how fast, and where.
The twist: broad “new skills” lift employment; explicit AI skills don’t—yet
One in ten postings in advanced economies now asks for at least one skill that wasn’t common a few years ago. In emerging markets that rate is roughly half as high, but rising. These demands concentrate in professional, technical, and managerial roles, and more than half sit inside the IT bucket. Where local markets absorb these new skills broadly—think accounting teams adopting data tools, operations managers using automation dashboards—employment doesn’t just hold; it grows. The IMF’s U.S. analysis finds that a one‑percentage‑point increase in the share of postings requiring new skills is associated with a 1.3 percent rise in local employment over the last decade. Wages move up alongside it. The lesson is that when skills diffuse across many jobs, firms add people to do more productive things.
But the curve bends when the skills are explicitly AI. Vacancies that demand AI skills pay a premium to the individual who lands them. At the regional level, though, a surge in AI-skill demand does not translate into broader employment gains, and in occupations highly exposed to AI with little complementarity, employment is 3.6 percent lower five years after AI skills appear. The pressure lands squarely on white‑collar, middle‑skill roles and, most worryingly, those just starting their careers. This is the first large-sample, cross‑country sign that the substitution story is not a speculative cliff but a gradual slope already under our feet.
Complementarity is the hinge
The IMF’s framing is refreshingly practical. AI isn’t a monolith; it either complements your tasks or substitutes for them. When the technology becomes a tool inside a role—analysts drafting faster with models, project managers monitoring workflows with intelligent copilots—it expands output and often headcount. When the tech becomes the role—routine drafting, summarizing, or pattern-based decisioning—the job shrinks even if the wage for the remaining roles rises. The paper separates exposure from complementarity to show that the same model that augments a designer can hollow out a junior compliance desk, and that managerial design choices determine which path a firm takes.
Translate this to the individual. A twenty‑something business analyst in Munich watches postings in their region add “LLM evaluation,” “workflow automation,” and “vector databases.” Their salary prospects look good if they pick up those tools. Their team prospects might not. If the team restructures around a few power users and a stack of agents, the number of analyst seats falls while the pay for the remaining ones climbs. The five-year time window in the data is the hard part: the labor market gives you a warning—but not a long one.
From anecdotes to architecture
This analysis isn’t stitched from headlines. It leans on millions of postings and worker profiles across 2010 to 2025, covering the United States, the United Kingdom, Germany, Denmark, Brazil, and South Africa, with local market analyses that link skill demand to employment and posted wages. The Fund pairs these near-term results with its earlier estimate that almost 40 percent of global jobs face AI-driven change, and it reframes the stakes beyond GDP: this is about how societies translate productivity into work, dignity, and purpose. In other words, no one gets to hide behind the idea that disruption is still hypothetical.
Policy gets new instruments, not just new slogans
To stop preaching and start prioritizing, the IMF introduces two lenses. The first, a Skill Imbalance Index, contrasts where demand for new skills is likely to outrun domestic supply. Countries with tight skill bottlenecks—Brazil and Mexico are mentioned, as is Sweden for specific domains—should push training, STEM pipelines, and when necessary, skilled immigration. Places where supply already exceeds demand—Australia, Ireland, Poland—need to accelerate firm adoption with innovation policy and access to finance. The second, a Skill Readiness Index, ranks countries on their capacity to produce both IT and non‑IT skills; Finland, Ireland, and Denmark surface as well‑positioned because they invested early in tertiary education and lifelong learning. These are not league tables for press releases. They are budget maps for labor ministries and central agencies deciding which levers to pull this fiscal year.
What this means right now
If you work in a white‑collar, middle‑skill role, your leading indicator is not a forecast about AGI; it’s the mix of skills in local job ads. When “new skills” in your function spread broadly, the region tends to add jobs and raise pay. When postings pivot to AI-specific requirements inside occupations that overlap with your daily tasks, the five‑year view tilts toward fewer seats. The safest move is to become the complement: learn the tools that expand your output and the judgment that the tools don’t replace. Mobility—across teams, firms, or cities—matters more than it used to, because the data show these effects are local.
If you run a company, the evidence hints at a design choice. Diffusing non‑AI “new skills” across functions correlates with headcount growth and wage gains—bigger pie dynamics. Concentrating on narrow AI hires without rethinking workflows risks consolidating work into fewer hands—smaller team, higher pay, thinner bench. The technology is the same; the org chart outcome isn’t.
If you set policy, the Fund’s advice is blunt: expand retraining and social protections now, and treat job postings as a leading indicator, not a curiosity. Align immigration with skill bottlenecks, subsidize firm adoption where supply is ready but demand lags, and redesign education so graduates arrive fluent in using, evaluating, and governing AI systems. The goal isn’t to slow AI; it’s to tilt it toward complementarity where possible and cushion the places where substitution is already underway.
The uncomfortable forecast—and the opportunity
The numbers aren’t apocalyptic; they are persistent. A few percentage points of employment loss in exposed roles, compounding over cohorts, is how middle-class erosion happens. The counterfactual is also in the data: regions that push broad skill adoption grow and pay better. The IMF just gave everyone—from finance ministers to frontline managers—a map. Whether this turns into widely shared gains comes down to execution within a narrow window. For readers of this blog, that window is personal. Watch the language of jobs around you. If AI skills start showing up in posts that mirror your tasks, you’re on a five‑year clock. Use it to move from being the replaced to being the one the system now depends on.
