New Belgian Study Confirms: Young Workers Risk Becoming the First Casualties of AI at Work

(Photo by Brooke Cagle) Across advanced economies, AI is moving from experiment to everyday infrastructure. Enterprises are rolling out chatbots, copilots and automation tools at scale. The first people to feel the real pressure are not senior experts, but young workers and administrative staff. It is something we also already reported about when we said that automation and AI are eating away at entry-level white-collar tasks.

A recent report by the Belgian institute High Council for Employment confirms this. Instead of treating that as a local anomaly, it’s better to see it as an early warning for what many labour markets are drifting towards.

Most of the data I will cite in this article comes from high- and upper-middle-income countries with good statistics. Many low-income countries however lack comparable AI and labour-market data, so the situation of young workers may differ and remain under-documented.

AI Adoption Is Accelerating Everywhere, Not Just in Belgium

In 2025, 20% of EU companies with at least 10 employees used some form of AI. That share almost doubled in just two years. Large firms lead the charge: more than two in five big companies already embed AI in their operations.

Similar patterns show up across OECD countries. Surveys compiled by OECD indicate that about a quarter of jobs sit in occupations with a high risk of automation once AI is taken into account. These are roles where a large block of tasks is routine, rule-based and therefore easy to codify.

Belgium fits right into that broader picture, just at a slightly higher speed. Its High Council for Employment estimates that roughly 43% of jobs there are “highly exposed” to AI. Managers and professional roles see AI mostly as a productivity booster while administrative workers face more direct substitution.

The important point: that basic structure – AI as a complement for higher-skilled office workers and a replacement for routine roles – appears in multiple national reports from the UK, Australia and others as well.

Where AI Helps and Where It Simply Replaces

Globally, AI is entering three broad categories of jobs, all of which we had already listed in our 2025 article on jobs AI would replace:

1. High-skill analytical roles: Think data-heavy management, software engineering, finance, design, legal drafting. Here AI automates repetitive parts of the work, surfaces insights faster and reduces search time. Most of the job still requires judgment, negotiation and domain knowledge. AI acts as a co-pilot here.
2. Customer-facing and care roles: Call centers, sales, healthcare, education. AI takes over triage, standard answers and documentation. But the core of the job, which includes empathy, persuasion, physical presence, responsibility, still rests with humans.
3. Administrative and clerical roles: Scheduling, data entry, standard emails, templated reporting, document formatting, invoice checking. This is where AI and other software tools can replicate an entire task flow with very little human input.

Research from the International Labour Organization (ILO) shows that AI threatens a particularly large share of administrative and clerical work. These are occupations that are also heavily female-dominated in many high-income countries. The technology absorbs a big slice of their value and goes beyond just ‘assisting’.

UK forecasts suggest up to three million low-skilled jobs could disappear by 2035, with trades, machine operation and administrative functions among the hardest hit. Australian projections point to around a quarter of jobs being at high risk of automation, again with office clerks and bookkeepers in the danger zone.

The pattern is quite consistent:

• High-skill knowledge roles get more leverage.
• Frontline care and service roles get new tools, but not full replacement.
• Administrative workers sit directly in the firing line.

The Belgian numbers show a high AI exposure for administrative staff with low “complementarity”, and display just one concrete national snapshot of this global trend.

Why Young Workers Are Hit First by AI Automation

Most young workers and their careers start at the bottom of a ladder built from routine tasks:

• checking data and correcting simple errors
• drafting standard emails and documents
• preparing basic reports
• doing background research and note-taking

These are exactly the tasks that generative AI handles well. Large language models write emails, summarise reports and generate templates in seconds. For employers under cost pressure, the temptation is obvious: let AI do most of the entry-level work and hire fewer young workers or juniors.

Several things follow from that:

• Fewer stepping-stone roles: when AI absorbs routine work, there are fewer positions in which young workers can learn by doing.
• Compressed learning time: young workers who are hired may see a narrower set of tasks, with AI inserted between them and the “real” work.
• Higher bar for entry: employers increasingly expect work-ready graduates who already master tools, domain knowledge and soft skills, and all this without first offering the old training ground of simple tasks. However, most of the young workers do not have this skills set readily available after graduating.

OECD analysis shows that workers in roles with high shares of easily automatable skills are disproportionately young and low-skilled. These are the people who traditionally gained experience by doing the work that AI is now taking over.

Young Workers Display Heavy AI Use, Weak AI Preparation

The whole shows quite a strange paradox: young workers use AI the most, but they are not always prepared to pilot or form it.

Across the EU, 63.8% of people aged 16–24 used generative AI tools in 2025 so a study by Eurostat shows. That is almost double the 32.7% share in the general population. The Eurostat reports shows that most of this usage is for personal tasks, study support and creative exploration.

At the same time, Europe’s “Digital Decade” strategy openly admits the skills gap they see with young workers and older workers: without new measures, only around 60% of adults will have at least basic digital skills by 2030, far below the 80% target. Advanced skills like data literacy, algorithmic thinking and AI governance are even less widely spread.

National snapshots show the same tension. In Belgium, for example, the High Council for Employment report shows that only a minority of 16- to 24-year-olds – future young workers – reach digital skill levels above the basic threshold, well below neighboring frontrunners. Young workers are confident users of apps and platforms, but less familiar with how systems are designed, where their limits lie, or how to audit their output.

The result shows a sharp imbalance:

• High everyday AI use (chatbots, image tools, homework helpers)
• Low structural understanding (how models work, where bias creeps in, how to check reliability)

In a workplace where AI is baked into core tools (office suites, CRM systems, coding environments) that gap matters a lot. Young workers (but also older workers) who only know how to “prompt” a system remain replaceable. Workers who understand when not to use AI, how to spot failure modes and how to redesign processes remain essential.

Gender: AI Pushes on the Weakest Link in the Labour Market

AI doesn’t hit everyone uniformly. As mentioned earlier in this article, women hold a large share of administrative and clerical roles in many economies, as a result they face a disproportionate risk of task automation.

The ILO estimates that, in high-income countries, occupations dominated by women are nearly three times more exposed to AI-driven transformation than male-dominated ones: 9.6% versus 3.5% of employment. Much of that difference comes from AI’s strength in office support tasks, secretarial work and back-office processing.

What Belgium’s High Council for Employment warns for administrative staff is, again, a local instance of a broader pressure point: if AI wipes out the routine parts of office work and employers do not open new, more qualified roles, female employment and financial independence take a hit in the exact segment of the labour market that was once seen as stable.

AI Hits Where There Are Few Shortages

Another pattern repeats across countries:

• Severe labour shortages in care, construction, technical trades and green-transition jobs.
• High AI risks in office support, basic accounting, routine legal and HR tasks, simple content production.

Belgium’s labour market illustrates this: construction and care are labelled “bottleneck professions”, however it are administrative roles that look more and more exposed to automation. UK, Australian and European projections all point to similar splits.

So the immediate threat is not mass unemployment, but rather a clear mismatch:

• jobs vanish in segments where AI can take over
• jobs remain unfilled where human labour is still crucial and physically demanding

For young workers and administrative workers, that means the challenge is not just “keep your job”. It is “move into a completely different type of job, often in another sector, and gain skills fast enough to be hired there”. The risk of underemployment is of course quite real here.

What Needs to Change: Four Concrete Priorities

The Belgian report offers just one concrete case. But global evidence from the OECD, ILO and others all points in the same direction as I said earlier. Taken together, they suggest four priorities for governments, education systems and employers.

1. Redesign Entry-Level Jobs

Instead of letting AI swallow all routine tasks, organisations can keep some of them deliberately as training material. That means:

• structuring junior roles as apprenticeships, with a planned progression for these young workers from simple to complex tasks
• pairing AI tools with human mentoring, so young workers see how decisions are taken and why prompts are adjusted
• making “explain what the AI did and where it might go wrong” a standard part of the job especially for young workers

These measures can keep the first rungs on the ladder intact, and keep young workers entering the labour market, even when AI is powerful.

2. Turn Administrative Workers into Process and Quality Experts

For administrative staff however, the realistic pathway is not to go “back” to pre-AI tasks. It is to move up the abstraction ladder:

• from typing invoices to designing and monitoring invoicing workflows
• from filling forms to checking AI output, spotting anomalies and handling exceptions
• from raw data entry to data quality and compliance roles

That requires funded, targeted reskilling programmes with recognised credentials, not just short internal workshops.

3. Make AI Literacy as Basic as Reading and Writing

Digital basics (office software, email, safe browsing) are no longer enough. As part of their Digital Decade agenda, the EU already set a target of 80% of adults with basic digital skills by 2030; right now, the trajectory points to around 60% if nothing changes. European Commission estimates that gap will persist without coordinated action.

A realistic baseline for “AI-ready” workers now includes:

• understanding what different AI systems actually do (prediction, generation, classification)
• knowing typical failure modes: hallucination, bias, outdated data
• being able to combine AI tools with traditional skills (statistics, logic, writing, domain knowledge)

STEM education matters, but so does strengthening critical thinking, data literacy and ethics in non-technical tracks.

4. Bake Inclusion and Gender Equality into AI Policy

Because AI affects female-dominated clerical roles so heavily, any serious AI labour-market policy has to integrate gender from the start. That includes:

• reskilling programmes specifically aimed at administrative workers, with childcare and financial support
• campaigns and pathways to bring more women into technical, care and green jobs where demand is growing
• monitoring systems that track AI-related job changes by gender, age and education level, not just by sector

Otherwise, AI will quietly deepen gaps that labour-market policy has tried to close for decades.

The Real Risk is a Lost Generation of Starters

Current evidence still doesn’t support the idea of an immediate, global “AI jobpocalypse”. AI-related layoffs are – for now at least – a small part of overall job cuts, and many sectors keep hiring. But the combination of trends is worrying in a different way:

• AI expands fastest exactly in the parts of work that used to train young workers.
• Administrative roles, a traditional entry route for women and non-graduate workers, are highly automatable.
• Digital and AI skills policies are not catching up with the speed of deployment.

Belgium’s High Council for Employment has put numbers on this dynamic for one country. Similar analyses from the UK, Australia and international organisations show that the same forces are at work elsewhere.

If policymakers and employers treat AI as just another efficiency tool, young workers and administrative staff will absorb most of the shock. If they treat AI as a reason to redesign entry-level work, reskill office staff and upgrade education fast, then those same groups can become the first real beneficiaries of the technology (instead of its first casualties).

---

---