Two things are true about AI and hiring in Asia at the same time, and the IMF has now measured both. The first is the uncomfortable one: employer demand for workers in AI-exposed occupations has fallen nearly 10 percent relative to less-exposed jobs since ChatGPT’s launch in November 2022. The second is the surprise: that decline has stopped growing — it has stabilised, even as AI systems became dramatically more capable.
That pairing comes from the IMF’s October 2026 Regional Economic Outlook for Asia and the Pacific, whose second chapter — “Labor Markets in Asia: Challenges and Opportunities in the AI Era” — analyses roughly 40 million online job postings across 19 Asian economies, tracking skills demanded by economy, industry and occupation, month by month through April 2026. As the Times of India’s summary of the findings put it, the data points to a shift in employer demand that does not suggest AI is simply eliminating jobs.
🔍 THE BOTTOM LINE
The most complete job-postings evidence yet assembled for Asia shows AI reshaping hiring rather than erasing work: a real but bounded dip in demand for exposed occupations, a decline that stopped worsening as AI capability kept climbing, and a growing counter-demand for workers who can prompt, build and integrate AI. Reduced hiring for some task profiles is still the stronger force, by the IMF’s own reading — but it is no longer the only one, and the offsetting opportunities are growing.
The Decline, Measured Properly
The IMF’s method matters, because job-postings studies are easy to do badly. The chapter compares hiring in more-exposed occupations (its example: writing code, where AI has become “surprisingly good”) against less-exposed occupations (its example: managing supplier relationships) — but it makes that comparison within the same economy, same industry and same month, and directly accounts for remote-work scope. Before 2022, hiring in the two groups moved in lockstep; little changed through mid-2023; the measurable weakening arrived in 2024 and 2025.
The IMF also flags the arithmetic of job postings honestly: they measure only new hiring, so the full weight of a hiring slowdown accumulates gradually over years. That is a reason for steady attention rather than either panic or complacency.
The Part Nobody Expected
Here is the finding that cuts against the run of alarm. The IMF lined up shifts in employer demand against external benchmarks for how long tasks AI can complete (citing METR’s 2026 capability estimates). When ChatGPT launched, AI could handle person-seconds — simple arithmetic. Little happened to hiring. By the time AI could reliably do person-hour tasks, such as finding and fixing bugs in a small code library, demand in exposed occupations sat nearly 10 percent lower.
Then the surprise: AI can now handle tasks that would take a person a full day, such as training a machine-learning model — and, in the IMF’s words in the chapter’s key passage, “the jobs picture has not worsened.” Demand remains below where it would otherwise sit, but the gap has stopped widening. Employers, the analysis concludes, still see a need for workers even in occupations where AI is useful.
Krishna Srinivasan, director of the IMF’s Asia and Pacific Department, framed the same result at the Bangkok press briefing on 14 October, as covered in the IMF’s own briefing record: negative effect on hiring in more AI-exposed jobs, employer demand down around 10 percent — but stabilised even as AI capability increased, “suggesting that AI is reshaping work more than eliminating it.”
The Offset: Skills That Did Not Exist in 2022
Why has the decline plateaued? The IMF built a measure of demand for what it calls emerging AI skills — prompt engineering, retrieval-augmented generation, building AI agents — skills that “meant little before 2023 and have since spread quickly.” Demand for those skills is rising fastest in exactly the occupations most exposed to AI. The chapter quantifies the adoption gap too: only about 6 percent of firms in Asia explicitly ask for AI skills in their postings, and at AI-adopting firms, demand for workers is growing faster than at firms that do not seek AI skills — including in exposed occupations.
The Hindu BusinessLine’s report on the chapter highlighted the same two-sided reading: AI can complement human capabilities even as it reduces demand for particular tasks, and the report flags uneven adoption as both a brake on displacement and a challenge for policy. CNBC-TV18’s coverage noted the finding — based on the 40-million-posting analysis — that the decline has stabilised even as AI capabilities have improved.
The honest caveat sits in the same pages: the IMF’s own summary of the two forces is that reduced hiring “remains more important” in the evidence to date, explaining the net decline in exposed demand — while the offsetting opportunities have “already” begun emerging and appear to be growing over time. Low-income economies in the sample show little measurable impact so far. Nobody should read a stabilised decline as a solved problem; it is a narrowing problem.
What It Means for New Zealand
New Zealand’s data has been tracing the same shape from the other direction. This site’s analysis of SEEK’s September employment report — record 4.2 percent of NZ job ads now mentioning AI skills while ads for the most automatable roles shrink — mirrors the IMF’s two-force picture: growth concentrated in people who work with AI, a real decline in the narrow band of most-automatable roles, and, per SEEK itself, cyclical factors alongside any AI effect. The trans-Tasman gap in AI-skill asks (NZ nearly double Australia) gets its first plausible macro explainer from a regional analysis like this: the Asia-Pacific AI boom is not a distant story for a labour market that lives on its trade.
Two further frames from our own archive: the IMF staff work mapping New Zealand’s AI exposure at about 60 percent of workers sits squarely in this report’s orbit, and the skills the chapter identifies as the offset — prompting, building, integrating — are the same halfway-overlap skills our coverage of the single-skill premium found commanding outsized pay. For NZ job seekers, the practical signal is unchanged but now better evidenced: the demand is not hiding, it is clustering.
FAQ
How many job postings did the IMF analyse? Roughly 40 million online job postings across 19 Asian economies, tracked month by month through April 2026, in Chapter 2 of the IMF’s October 2026 Regional Economic Outlook for Asia and the Pacific.
Did AI reduce hiring in Asia? Yes, measurably: employer demand in more AI-exposed occupations fell by nearly 10 percent relative to less-exposed occupations after ChatGPT’s November 2022 launch, within the same economy, industry and month.
Has the decline continued as AI got better? No — and this is the report’s most striking finding. The decline stabilised even as AI capability advanced from person-second tasks to day-long tasks. The IMF attributes the plateau to growing employer demand for emerging AI skills.
What are “emerging AI skills”? Skills that barely existed before 2023: prompt engineering, retrieval-augmented generation (RAG), and building AI agents. Only about 6 percent of Asian firms request AI skills in postings today, and AI-adopting firms show faster demand growth than non-adopters.
What does this mean for New Zealand? NZ’s own job-ad data shows the same pattern: record AI-skills mentions alongside a decline only in the most automatable roles. A regional Asian analysis matters for NZ because its labour market, exporters and offshore tech services are tied to the same boom.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
