Why starting a career is becoming harder with AI
AI generated
In my last blog post, I explored a central insight from recent research: AI does not (yet) destroy jobs – but it reshapes them.
However, there is one group for whom the picture looks considerably less optimistic: early-career professionals. Here are some reasons why:
- The disappearing ‚first steps‘
For decades, organisations relied on a relatively stable pattern: Entry-level employees handled rather routine tasks like research, documentation, data preparation etc. Doing this, over time, they developed judgment, contextual understanding, and professional intuition.
Thus, this was never just “simple work.” It was learning infrastructure.
Today, this infrastructure is eroding. Recent studies (e.g. McKinsey, Revelio, Stanford) show that precisely these tasks are increasingly being taken over by AI systems and that the demand for entry-level roles is declining sharply. While demand for experienced workers is holding steady or even increasing. - A fundamental shift
The challenge is not losing jobs. The challenge is not getting in. Entry-level roles are becoming fewer, more demanding, and less “forgiving”.
In other words: The threshold for becoming productive is rising. - From learning-by-doing to learning-by-design?
Traditionally, organisations relied on a simple mechanism: People learn through doing – even if the work is imperfect. AI disrupts this mechanism in two ways:
Task removal: The “practice field” disappears.
Task delegation: Instead of doing the work, juniors supervise AI output.
And here lies a paradox. Experimental research shows that heavy reliance on AI can actually impair skill development, particularly for novices: weaker conceptual understanding, reduced ability to diagnose errors, and less deep learning.
So we are asking early-career professionals to start at a higher level with fewer opportunities to learn while relying on tools that may undermine their learning. - The missing piece: Seniors because they work at home
If entry-level work becomes more complex, the obvious answer is: more support from experienced colleagues. But this is exactly where the second major trend comes into play: remote and hybrid work.
Research (e.g. Emma Harrington, University of Virginia and again Emma) shows:
Remote work makes it harder to train and mentor junior employees. Physical proximity significantly increases feedback and learning. Career progression slows when mentoring is reduced.
There is even evidence suggesting that a large share of rising youth unemployment can be linked to the spread of remote work.
From a knowledge perspective, this is not surprising. Learning in organisations is not just about formal training, explicit knowledge, and documented processes. It is about informal feedback, observation, asking “quick questions”, being pulled into conversations. In short: tacit knowledge transfer. And tacit knowledge does not travel well over Teams. - Double disruption
If we combine both developments, a problematic dynamic emerges:
AI removes beginner tasks, raises entry requirements, speeds up workflows. Remote work removes informal learning, reduces access to mentors, slows down capability building. - A Knowledge Management blind spot
From a KM perspective, this should ring alarm bells. Many organisations are currently focusing on AI tools, automation, efficiency gains, and knowledge repositories.
But they are overlooking more fundamental questions: How do people develop the capability to work with knowledge in the first place? Where do juniors practice? Where do they make mistakes? Where do they receive feedback? Where do they observe experts at work?If these mechanisms are not redesigned, we risk creating a generation of workers who are highly tool-enabled but insufficiently experienced. - Rethinking the entry phase
Based on the emerging evidence, three areas seem critical:
Deliberate learning design (structured learning tasks, well designed progression paths, combination of AI use with reflection)
Reinvestment in mentoring with structured check-ins, conscious feedback loops, and visible expert work.
Hybrid thinking (not just hybrid working), which means that if proximity matters for learning, then presence must be intentional. Thus, not “come to the office sometimes” but „come together when learning happens“.
So – who should be worried now?
In the last post, the answer was: It depends on the tasks. For early-career professionals, the answer becomes sharper: You should be concerned if your learning depends on routine tasks, passive observation, and implicit guidance. Because all three are under pressure.
Final thought
The first blog post asked: Who is afraid of AI? Who should be?
This follow-up suggests a different question: Who will still be able to learn how to work? Because the real long-term risk is not job loss. It is the quiet erosion of the pathways that create expertise. And that is, at its core, a knowledge management challenge.
