How AI is reshaping the job market: which roles are growing, which are shrinking

Customer service: Klarna announced reducing its customer service headcount from 2,300 to 1,600 (a cut of 700) after deploying AI; AI now handles 70% of all customer interactions. This is one of the most concrete and most-cited examples in the industry.

Which jobs are growing?

According to World Economic Forum projections, 170 million new roles will be created globally by 2030 while 92 million roles disappear — a net gain of 78 million jobs. Roles like AI prompt engineering, machine learning specialization, and AI ethics officer are becoming standard positions. Entirely new job descriptions are also emerging on the software engineering and operations side.

The group hit hardest: entry-level workers

Perhaps the most striking finding: global entry-level job postings have fallen 29% since January 2024. The logic is simple — routine, repeatable tasks tend to be concentrated in first-rung career jobs, and that's exactly the kind of work AI does best. As a result, experienced workers stay relatively safe while people starting their careers face a much harder entry into the market.

Which skills actually get priced in?

Read the language of the postings and it is clear employers are not paying for "knowing AI" — they are paying for the ability to attach AI to a business outcome. The skills commanding the biggest premium fall into three groups: being able to verify a model's output (domain knowledge), being able to automate a process (data and integration knowledge), and being able to manage the risk (legal, compliance, security).

That does not mean traditional professions are finished. If anything, domain experts learning the tools are moving faster than tool experts learning the domain. An accountant automating a reconciliation workflow with AI gets there sooner than an engineer learning accounting rules.

At the same time, "writing prompts" is not a career on its own. Prompt engineering, treated as a separate title in 2024, has largely been absorbed into ordinary job descriptions by 2026. Folding the skill into your existing role is far more realistic than hunting for a dedicated position.

How to position your career now

The most fragile roles are rule-based, standardised jobs whose inputs and outputs are entirely digital. There is a simple test for how exposed a job is: if you can write down what you do step by step and those steps are the same every time, the work is open to automation.

Roles that involve ambiguity, accountability and negotiation between people are far better insulated. So the practical strategy is to deliberately hand the repetitive part of your job to tools and move the freed-up time toward judgement, relationships and oversight.

The picture is harder for new graduates: with entry-level postings shrinking, the classic "learn on the job as a junior" path is narrowing. What works here is visibility through a portfolio — small projects that solve real problems are more convincing to most employers than another course listed on a CV. The market itself is also shifting: a meaningful share of the roles being posted today did not exist five years ago, which makes investing in transferable skills a safer bet than chasing a job title.

What does this data actually mean?

Three practical takeaways:

1. "Will AI take my job" is the wrong question. The better question is: "Which part of my job is routine and repeatable, and which requires context, judgment, or relationships?" The routine part shrinks; the judgment-requiring part gets more valuable.

2. AI skills are no longer a "bonus" — they're a baseline expectation. The 62% wage premium suggests this is transitioning from a differentiator into a minimum expectation.

3. The entry-level squeeze is also a risk for companies. While it looks like short-term cost savings, the shrinking pool of experienced workers who'll rise into senior roles is a long-term cost most companies haven't accounted for yet.

Bottom line: AI isn't destroying the job market — it's reshaping it. But the burden of that reshaping isn't distributed evenly, and right now it's falling hardest on young, entry-level workers.