Last updated: October 12, 2026
The IMF estimates AI will touch about 40% of jobs worldwide and 60% in advanced economies, but roughly half of those exposed jobs gain from it, so your real decision is which roles get augmented, downgraded, or cut.
Kristalina Georgieva said it out loud on Monday in Bangkok, at the IMF and World Bank annual meetings. AI is “a slow-moving tsunami, but [a] tsunami nonetheless.” Most of the coverage stopped at the wave. I’d rather talk about the word in front of it.
Slow is the part a hiring plan can actually use. A fast shock forces you to react. A slow one gives you time to decide, seat by seat, what each open role is going to look like in 2028, and that is a very different exercise from freezing everything until the economists agree. I run searches with engineering and infrastructure leaders through our IT staffing services practice, and since the weekend the question has changed from “is this real” to “which of my reqs is it actually about.” Good question. It has an answer.
What Georgieva Said in Bangkok, and What She Didn’t
Start with what’s new, because less of it is new than the headlines suggest. Much less.
The 40% and 60% figures are not fresh estimates. They come from an IMF staff discussion note published in January 2024, which Georgieva summarized on the IMF blog at the time: “almost 40 percent of global employment is exposed to AI,” and in advanced economies “about 60 percent of jobs may be impacted.” She has repeated the tsunami line since May 2024. What changed this week is the tone and the evidence behind it. According to Bloomberg’s report from the Bangkok panel, she said young workers in AI-exposed fields have already seen hiring slow, that AI is pushing up some low-paying work in sectors like construction (“not a story to celebrate”), and that nobody can “imagine what exactly the world of work will be five years from now.”
Then growth. The calendar version of this story says AI is holding global growth near 3%. Close enough, with one caveat. The current number on the books is the IMF’s July 2026 World Economic Outlook Update, which projects 3.0% global growth for 2026 and 3.4% for 2027, with the war-driven energy shock “partly offset by accelerated demand-driven momentum in the global technology cycle.” U.S. growth sits at 2.3%. The October outlook is due Tuesday, so if you’re reading this later in the week, check whether it moved.
Put those together and you get an odd picture. AI spending is propping up output while it quietly thins out the entry points into the workforce. Both things are true at once. A CFO looking only at GDP will think the coast is clear.
The construction remark deserves its own note. When I wrote about the Broadcom and Anthropic chip financing last week, the first hiring pressure I expected was electricians and commissioning crews, not software engineers. Georgieva is describing the same thing from the other end. The jobs AI creates first are physical, and plenty of them pay less than the desk jobs it’s squeezing.
AI Impact on Jobs: Exposure Is Not a Pink Slip
AI exposure, in the IMF’s usage, means a job contains tasks that AI can perform or assist with. It does not mean the job disappears. The IMF’s own note splits exposed jobs roughly in half, with one half likely to gain productivity from AI and the other half facing lower labor demand, lower wages, or in extreme cases elimination.
That split gets lost every time the 60% number travels. Here is the full set the IMF published.
| Economy group | Share of jobs exposed to AI | What that means for a U.S. employer |
|---|---|---|
| Global | About 40% | The headline figure, and the least useful one for you |
| Advanced economies (including the U.S.) | About 60% | Your workforce, since cognitive desk work is concentrated here |
| Emerging markets | About 40% | Matters if you run offshore delivery centers |
| Low-income countries | About 26% | Lower exposure because AI arrives later, not because those jobs are safer |
| Exposed jobs likely to benefit | Roughly half | These seats get more productive and often harder to fill |
Source: IMF Staff Discussion Note 2024/001 and Georgieva’s January 2024 summary. Some outlets quoted 23% for low-income countries this week; the published figure is 26%.
So take 60% of your headcount, cut it in half, and you have the slice that’s genuinely at risk. Still large. Not apocalyptic. And it’s spread across your org chart unevenly enough that a blanket response will be wrong for most of the seats it touches.
Augmented, Downgraded, Eliminated: Sorting Your Own Req List
Georgieva used three verbs on Monday. Augmented, downgraded, eliminated. They map almost perfectly onto the decisions a hiring manager actually makes, which is more than I can say for most AI frameworks I’ve been sent this year.
Pull every open and planned req for the next 18 months. Put each one in a bucket. Do it per role, not per department, because a single finance team will usually have all three.
| Bucket | Typical seats we see in it | What changes in the req |
|---|---|---|
| Augmented | Senior software engineers, cloud architects, security engineers, FP&A managers | Same count or fewer, higher bar, comp holds or rises |
| Downgraded | Tier 1 SOC analysts, QA testers, staff accountants, help desk | Scope narrows, title and band come under pressure |
| Eliminated | Manual data entry, first-pass invoice matching, scripted chat support | Req closes or merges into a different role |
| Not exposed | Field techs, electricians, data center commissioning, lab work | Demand rises if you’re near the AI build-out |
Augmented seats are where people get surprised. A senior engineer who ships twice as much with Copilot or Claude Code isn’t a cheaper hire. She’s a more valuable one, and every competitor running the same math wants her too. We filled a staff platform engineer role in September for a SaaS company in Irvine where the hiring manager cut the team plan from four engineers to three and raised the band on all three by $22K. That’s what augmentation looks like on a budget sheet. Fewer seats, pricier seats.
Eliminated seats are the easy part, honestly. Everyone already knows which ones they are. If a role is mostly keying invoices from PDFs into NetSuite, nobody needed the IMF to tell them.
The middle bucket is the problem.
The Downgrade Bucket Nobody Budgets For
“Downgraded” is the IMF word that got the least airtime this week, and it’s the one I’d lose sleep over. A downgraded job still exists. Someone still sits in it. But the AI handles the parts that used to justify the pay band, so what’s left is narrower, more procedural, and easier to swap out.
We’re watching this happen to tier 1 security analysts right now. An AI triage layer closes most of the low-severity alerts before a human sees them, so the analyst who used to spend six hours a day on alert queues spends two, and the remaining four hours are either escalation work (which needs a more experienced person) or nothing. Companies respond in one of two ways. Some fold the role upward and hire fewer, better analysts. Others keep the title, keep the headcount, and quietly freeze the band for three years. The second group loses their best juniors to the first group. I covered the security side of this in more detail in whether AI will replace cybersecurity jobs, and the short version still holds: the work changes faster than the org chart does.
Three things go wrong in a downgraded seat if nobody plans for it. The job description stops matching the job, so you hire for skills the person never uses. The comp band drifts out of market in one direction or the other. And the role stops being a training ground, because the tasks that used to teach a new hire the business are now done by software.
That last one connects straight to the youngest people on your payroll.
The First Rung Is Already Moving
Georgieva said young workers in exposed fields are already seeing hiring slow. The data agrees, and it’s been agreeing for a while.
Stanford’s Digital Economy Lab, working from ADP payroll records, found that workers aged 22 to 25 in the most AI-exposed occupations saw a 16% relative decline in employment after controlling for firm-level shocks, while experienced workers in the same occupations held steady (Brynjolfsson, Chandar and Chen, November 2025 version). Their monthly tracker, last updated September 23, shows early-career employment in the most exposed quintile contracting about 4.5% a year for women and 2.5% a year for men since November 2022. The New York Fed puts the recent-graduate unemployment rate at about 5.6% for the second quarter of 2026, with underemployment at 42%, against an overall unemployment rate of 4.2% in the BLS September jobs report.
Read that gap slowly. A 22-year-old with a degree is more likely to be out of work than the average American. That used to be backwards.
I’m not going to re-argue the case for keeping a junior pipeline here, because we made it at length in our piece on why smart employers are still hiring new grads, and nothing in Bangkok changes it. If your junior engineering pipeline leans on international graduates, the proposed $70K OPT fee would squeeze the same rung from another side. The IMF’s framing adds one thing worth saying. If AI takes the entry-level tasks first, and Georgieva said in Davos in January that it does, then the downgrade bucket and the junior pipeline are the same problem. Strip the learning tasks out of a junior role and you haven’t just made the job smaller. You’ve removed the way your 2030 senior engineers were going to learn the job.
Some teams are fixing this deliberately. One client, a mid-market medical device company, rebuilt their associate QA role this summer so the AI does the regression runs and the associate’s first six months are spent pairing on test design with a senior. Same headcount. Different job. It took them about three weeks of a director’s time to rewrite, and they say it was worth it. I believe them, though it’s early.
Why “Slow” Is the Word That Matters for a Two-Year Plan
A tsunami you can see coming for five years is mostly a planning problem.
The mistake I expect to see this quarter is companies treating a five-year structural shift like a two-week news cycle, cutting or freezing roles because of a forecast instead of because of anything that has actually happened in their own workflow. The IMF itself says it can’t picture the world of work five years out. Your CFO can’t either. Robert Ardell’s Q4 tech job market forecast already shows where that kind of reflexive freeze is landing this quarter. Locking in a permanent headcount reduction on that basis means betting a fixed cost on a guess, and it’s a guess the people who wrote the 60% figure won’t make themselves. I wouldn’t either.
What I’d do instead is match the commitment to the certainty. That’s the whole idea. Augmented roles, where you know the work is growing, can go to direct hire now, before competitors finish their own math and start bidding. Downgraded roles, where you don’t yet know what the job will look like in 2028, are a reasonable place for contract staffing or contract-to-hire, so you can watch how much of the work actually moves to software before you commit to a band. Eliminated roles you just stop backfilling, and you tell the people in them early.
Speed still matters on the augmented side. Our average time-to-hire for IT roles is 17 days, and for the AI infrastructure and senior platform seats it runs longer, because those candidates are juggling multiple offers. We wrote about that divergence in the senior SWE glut and the AI infrastructure drought. Waiting to see what the October outlook says won’t make those people any cheaper. If anything, the reverse.
Before Anyone Rewrites the Headcount Plan
Did the IMF just say 60% of American jobs are going away?
Exposure and job loss are different things, and the IMF’s 60% figure measures only exposure, meaning jobs in advanced economies with tasks AI can perform or assist with. The IMF’s own research expects roughly half of those exposed jobs to benefit through higher productivity rather than lose ground.
Is the 60% number new this week?
The 60% number dates to January 2024, when the IMF published Staff Discussion Note 2024/001, and Georgieva has repeated it at Davos and elsewhere since. What was new in Bangkok was her emphasis on young workers already seeing hiring slow and on low-paying jobs growing in sectors like construction.
So which roles get downgraded first?
Roles built around high-volume, rules-based tasks with a human check at the end. Tier 1 security alert triage, manual QA regression, first-pass accounts payable, and scripted support are the ones we see changing fastest in client job descriptions this year.
We already froze junior hiring. Mistake?
Probably, if the freeze is open-ended. A pause while you redesign what a junior role does is sensible. A freeze with no redesign behind it means you’ll be short of mid-level people in three or four years, and buying them back on the open market costs more than training them would have.
Does AI growth near 3% mean hiring will pick up?
Not evenly, because a lot of that lift is AI capital spending, which hires electricians, data center builders, and a small number of very expensive engineers rather than broad white-collar headcount. The IMF’s July outlook has global growth at 3.0% for 2026 and U.S. growth at 2.3%, and the October update lands this week.
How fast can KORE1 fill an AI-augmented senior role?
Seventeen days is our average time-to-hire across IT roles. Senior AI infrastructure and platform searches usually take longer than that because the candidates have options, so starting before the budget cycle closes is the single biggest thing that shortens them.
Monday Morning, Not 2031
The IMF gave everybody a long-range forecast this week. Fine. Your hiring plan runs in quarters.
Take the req list. Sort it into Georgieva’s three buckets plus a fourth for roles AI doesn’t touch. Rewrite the downgraded ones before you post them again, because a job description written in 2023 for a role that’s half automated in 2026 will attract the wrong people and pay them the wrong amount. Put the augmented seats on a faster track than feels comfortable. And don’t let anyone close the junior pipeline because of a headline about a wave that, by the IMF’s own description, is still a long way offshore.
If you want a second set of eyes on the sort, talk to a recruiter on our team. We do this exercise with clients every week now, and the answer is rarely the one the org chart suggests.
