Is There an 18-Month AI Jobs Cliff?

Some technology leaders say AI agents will replace much of office work within a year or two. The evidence so far shows real change in particular jobs and companies, but not a cliff: across the US labour market, researchers at Yale found no discernible disruption in the 33 months after ChatGPT was released.

That gap between warning and data matters for anyone deciding what to do now: a worker choosing what to learn, a manager planning a team, or a government designing a safety net.

Where does the 18-month warning come from?

The warning rests on what leading companies say and do.

• Klarna announced in February 2024 that its AI assistant had handled 2.3 million conversations, two-thirds of its customer service chats, in its first month, doing the equivalent work of 700 full-time agents.
• IBM's chief executive, Arvind Krishna, told Bloomberg in 2023 that hiring for back-office roles would be paused or slowed. Of about 26,000 such roles, he could “easily see 30%” replaced by AI and automation over five years, roughly 7,800 jobs.
• Nvidia's chief executive, Jensen Huang, told the World Government Summit in Dubai that children no longer need to learn to code, because AI will do it.

Forecasts point the same way. Goldman Sachs estimated that the equivalent of 300 million full-time jobs could be exposed to automation, and that about two-thirds of US occupations are exposed to some degree.

Is AI actually causing mass job losses?

Not yet, according to the best data available. The Budget Lab at Yale compared how fast the mix of US jobs has changed since ChatGPT’s release with earlier waves of technology, such as computers and the internet. It found no discernible disruption of the broader labour market in the first 33 months. Its authors stress that their work does not predict the future and that wider effects may take longer to appear.

Company stories also look less simple a year on. In 2025 Klarna’s chief executive told Bloomberg that putting cost first had led to “lower quality”, and the company began recruiting human agents again, while still describing itself as AI-first.

Exposure is not the same as replacement. The IMF estimates that almost 40% of global employment is exposed to AI, but expects about half of the exposed jobs in advanced economies to gain from AI rather than lose out. The World Economic Forum’s Future of Jobs Report 2025 expects 170 million jobs to be created and 92 million displaced by 2030, a net gain of 78 million, and expects 39% of workers' existing skills to change.

Who is most at risk?

Analysis, from the sources above: the roles most exposed are routine cognitive tasks that follow clear rules: first-line customer service, data entry, standard document handling and some back-office administration. The WEF lists data entry clerks and bank tellers among the fastest-declining roles. Jobs that combine judgement, relationships and physical presence are less exposed.

The risk is uneven. A sudden change for one occupation in one company can be a cliff for the people in it, even when national figures look calm.

Can a safety net like basic income help?

The IMF’s advice for every economy is that social safety nets and retraining for workers exposed to AI are crucial. A universal basic income is one option among several, alongside unemployment insurance, wage insurance and funded retraining.

Evidence on a guaranteed income is growing. In OpenResearch’s study, people who received $1,000 a month for three years worked 1.3 fewer hours a week on average and were 2 percentage points less likely to be employed than the control group. Whether that is a cost or a benefit depends on what people did with the time; the study’s full results are the place to judge. We look at the trials in Universal Basic Income and AI: What the Trials Show.

What should workers, employers and governments do now?

1. Workers: learn to use AI tools in your own job, and build the skills they do not replace: judgement, checking AI output, working with people.
2. Employers: test before you cut. Klarna’s reversal shows the cost of replacing people faster than quality allows. Keep humans responsible for decisions and for customers who need them.
3. Governments: watch the labour data monthly, strengthen unemployment insurance and retraining now, and pilot income support before it is needed in a hurry.
4. Everyone: be wary of precise deadlines. Neither “18 months” nor any other date comes from labour data.

For the governance side of moving fast with AI, see Does Cutting AI Governance Really Save Money?.

Frequently asked questions

Is AI causing mass unemployment yet?

Not in the data so far. The Budget Lab at Yale found no discernible disruption of the broader US labour market in the 33 months after ChatGPT’s release. Some companies have cut or paused hiring in specific roles.

How many jobs could AI affect?

Goldman Sachs estimated in 2023 that the equivalent of 300 million full-time jobs worldwide could be exposed to automation. The IMF estimated that almost 40% of global employment is exposed to AI, though exposure does not always mean replacement.

Will AI create jobs as well as remove them?

The World Economic Forum’s Future of Jobs Report 2025 expects 170 million jobs to be created and 92 million displaced between 2025 and 2030, a net gain of 78 million, with large shifts in the skills each job needs.

What happened when Klarna replaced customer service staff with AI?

In 2024 Klarna said its AI assistant did the work of 700 full-time agents. In 2025 its CEO said the focus on cost had led to lower quality, and the company began recruiting human agents again.

References

1. Klarna (2024), Klarna AI assistant handles two-thirds of customer service chats in its first month — 2.3 million conversations, two-thirds of customer service chats, the equivalent work of 700 full-time agents.
2. Fortune (2025), Klarna plans to hire humans again — The CEO told Bloomberg that the cost-focused AI approach led to lower quality and the company was recruiting humans again.
3. Bloomberg (2023), IBM to Pause Hiring for Back-Office Jobs That AI Could Kill — About 26,000 back-office roles; the CEO could see 30%, roughly 7,800, replaced by AI and automation over five years.
4. Tom's Hardware (2024), Jensen Huang says kids shouldn't learn to code — Nvidia’s CEO at the World Government Summit in Dubai.
5. Goldman Sachs (2023), Generative AI could raise global GDP by 7% — The equivalent of 300 million full-time jobs exposed to automation; about two-thirds of US occupations exposed to some degree.
6. IMF (2024), Gen-AI: Artificial Intelligence and the Future of Work, Staff Discussion Note SDN/2024/001 — Almost 40% of global employment exposed; social safety nets and retraining are crucial.
7. World Economic Forum (2025), The Future of Jobs Report 2025 — 170 million jobs created and 92 million displaced by 2030; 39% of workers' skills expected to change.
8. The Budget Lab at Yale (2025), Evaluating the Impact of AI on the Labor Market: Current State of Affairs — No discernible disruption of the broader labour market in the 33 months after ChatGPT’s release.
9. OpenResearch, Unconditional Cash Study — Evidence on what a guaranteed $1,000 a month does to work and wellbeing.