When AI Takes Over Entry-Level Work: How Will Tomorrow’s Experts Learn?

TL;DR

• Entry-level jobs are learning jobs. Early-career employees don’t just complete tasks. They build experience, develop domain expertise, and learn how to exercise judgment.

• AI is changing that learning stage. Many traditional junior-level tasks, from research and analysis to first drafts, can now be automated. That also means valuable learning opportunities may disappear.

• Today’s entry-level crisis could become tomorrow’s talent problem. If fewer people have the opportunity to gain hands-on experience, companies may eventually face a shortage of experienced specialists and leaders.

• Workplace learning will need to become more intentional. Mentoring, expert–novice relationships, shadowing, apprenticeship-style learning, and peer learning become more important when development no longer happens naturally through day-to-day work.

• AI literacy and human capabilities go hand in hand. Employees need to know how to use AI effectively, but they also need the expertise, context, and judgment required to evaluate its output.

• Not everything that can be automated should be automated immediately. Companies need to decide where AI should create efficiency and where certain tasks should remain valuable learning opportunities.

• The bigger question: If AI takes over some of the work people traditionally learned from, how do we make sure we are still developing the experts we will need tomorrow?

Breaking Into the Workforce Is Getting Harder, Especially in Knowledge Work

Not long ago, I spoke with a recent graduate looking for his first full-time job. On paper, he brings exactly what companies say they want: a strong university degree, digital skills, some practical experience, and a very natural ability to work with AI.

And yet, finding that first job has been much harder than he expected. Many of the tasks companies traditionally hired junior employees to handle can now be at least partially automated. At the same time, many job postings ask for experience that people are normally supposed to gain in their first or second role.

At one point, he essentially asked me: How am I supposed to gain experience if nobody is willing to give me the opportunity to gain it? That question has stayed with me. Because I think it points to a much bigger issue than the challenges individual graduates are facing.

Indeed data from Germany shows just how significant the shift has become. Entry-level job postings have fallen sharply since 2022 and, by July 2026, stood at just 46% of their 2019 level. The decline has been particularly pronounced in knowledge-intensive fields such as marketing, project management, and technology.

At the same time, Indeed makes an important distinction: this trend began before ChatGPT entered the mainstream. AI is part of the picture, but it would be too simplistic to treat it as the sole cause. Regardless of what is driving the decline, I don’t think fewer entry-level jobs are only a problem for young people trying to start their careers. Over time, they could become a serious strategic problem for employers as well.

Entry-Level Jobs Are More Than Cheap Labor

One thing strikes me whenever I read discussions about disappearing entry-level roles: junior employees and AI are often compared almost exclusively through the lens of speed, productivity, and cost. The argument goes something like this: Why hire a junior employee if AI can complete the same task faster, more accurately, and at a lower cost? I think that view is far too narrow. And in the long run, it could be dangerous.

If we see early-career employees primarily as inexpensive capacity, we overlook one of the most important functions entry-level work has always served. An entry-level role traditionally does two things at once: it produces results and gives someone the opportunity to build experience.

Most conversations about automation focus heavily on the first and barely acknowledge the second. But if companies optimize only for speed and cost today, they may discover a few years from now that they no longer have enough experienced specialists and leaders to move into more complex roles.

When AI Takes Over Junior-Level Tasks, Where Does Experience Come From?

Think about how many people have traditionally learned. A junior employee prepares an analysis. A more experienced colleague reviews it, points out mistakes, explains what matters, and asks questions. The junior tries again. Over time, the tasks become more complex, and the level of responsibility increases.

Now imagine a different process. AI produces the initial analysis. A senior employee reviews the output, uses it, adjusts it, or asks the system to generate another version. The junior employee may no longer be part of that process at all.

And that matters, because early-career professionals learn through doing: smaller assignments, observation, feedback, repetition, mistakes, context, and gradually increasing responsibility.

Over several years, they develop judgment in their field. At least today, that kind of judgment is still very difficult to replace with AI. That is why learning by doing matters so much. It is one of the ways organizations develop the experienced specialists and leaders they will rely on later.

If companies hand every basic entry-level task over to AI, where are people supposed to gain that foundational experience? And if we think back to our own careers: weren’t some of the lessons that stayed with us the longest the ones we learned by getting something wrong? What happens when we remove those opportunities to make mistakes from the process altogether?

Today’s Entry-Level Crisis Could Become Tomorrow’s Talent Gap

At first, the decline in entry-level jobs appears primarily to be a problem for recent graduates. And of course it is frustrating to spend years earning a degree only to find that there are fewer companies willing to give you the space to learn, make mistakes, and build experience.

From an employer’s perspective, a smaller pool of early-career talent may not seem particularly alarming right now either. Especially in a difficult economic environment, when automation offers an increasingly affordable alternative. But I believe companies may feel the consequences of this development much more clearly in five or ten years.

If fewer organizations develop junior employees into experienced professionals, the pipeline narrows. There will be fewer people reaching the mid-career level and, eventually, fewer senior experts. Companies will have fewer internal successors for critical roles and become increasingly dependent on external talent to fill them. In other words, the entry-level hiring decisions companies make today could shape the expertise available to them years from now.

Workplace Learning Will Need to Become Much More Intentional

I don’t think many organizations realize just how much learning happens during an ordinary workday. Over the years, informal learning structures have developed within and across teams without anyone deliberately designing them.

You sit next to someone more experienced and pick up how they approach a problem. You overhear conversations. You watch people debate a decision. You see why one option is chosen over another. Or you simply have someone nearby you can turn to with a quick question.

This is also why I often come back to the 70-20-10 model of learning. Formal training matters, but a huge part of professional development happens through experience and interaction with other people. As AI increasingly individualizes work or takes over certain tasks entirely, some of that informal learning system could start to disappear.

A junior employee can ask an AI assistant a question and receive an answer within seconds. But an answer is not the same thing as a conversation. There is no colleague who shares the rationale behind their recommendation, challenges the question itself, or explains why the obvious solution may not work in this particular organization.

We have long treated mentoring, knowledge transfer, and expert–novice relationships as things that simply emerge naturally at work. I’m not sure we can continue to assume that they will. Companies may need to become much more deliberate about creating the human relationships through which learning happens.

How Companies Can Create New Learning Paths for Early-Career Talent

1. Redesign Junior Roles Instead of Simply Eliminating Them
   Companies should not only ask, "Which tasks can AI take over?" They should also ask: Which tasks do early-career employees need in order to develop judgment? Shell offers an interesting example. Its learning paths for early-career employees have been designed so that people first frame the problem themselves, test assumptions, and develop an initial analysis before turning to AI. That keeps AI within the learning process without allowing it to replace it.

2. Treat Mentoring as Learning Infrastructure, Not Just an Employee Benefit
   Mentoring should increasingly be seen as part of an organization’s learning infrastructure rather than an optional perk. When traditional learning tasks disappear, early-career employees need intentional access to experienced colleagues who can provide context, offer feedback, and share the kind of knowledge that rarely exists in a manual or database.

3. Build Expert–Novice Relationships Deliberately
   Companies also need to create stronger relationships between experienced employees and those still building expertise, regardless of whether those relationships are part of a formal mentoring program. Early-career employees need regular access to people who can help them understand not only what to do, but how decisions are made, why certain approaches work, and what experience has taught them over time. What once happened almost automatically during everyday work may increasingly need to be designed on purpose.

4. Rethink Shadowing and Apprenticeship for Knowledge Work
   Shadowing should not be limited to observing how someone completes a task. Especially in knowledge work, early-career employees can benefit from joining experienced colleagues in meetings, complex projects, and decision-making processes. The real value lies in seeing the thinking behind the outcome. Why was one approach chosen over another? What trade-offs were considered? Which signals did an experienced colleague notice that a newcomer might have missed? For someone at the beginning of their career, that context is incredibly valuable.

5. Don’t Put AI Between the Employee and the Learning Task Too Early
   Just because AI can complete a task faster does not automatically mean it should take over that task. A task can be inefficient and still be valuable for learning. Research, analysis, and first drafts, for example, help people learn how to structure information, identify patterns, challenge assumptions, and form their own judgment. If AI does all of that from day one, we need to ask what capabilities the employee is no longer developing themselves.

6. Add Peer Learning to the Mix
   Early-career employees do not only learn from senior colleagues. Learning circles and peer groups can create valuable spaces where people compare experiences, solve problems together, and share how they use AI and adapt their workflows. They also create an environment where people can openly ask questions, test ideas, and admit when they are unsure. The World Economic Forum makes a similar point in its discussion of reskilling: preparing people for changing work cannot simply mean providing more formal courses. On-the-job experience, informal learning, and exposure to new projects remain central to how people actually develop new capabilities.

Not Every Task That Can Be Automated Should Be Automated

For me, this is ultimately not a question of whether companies should use AI. They will, and in many areas they absolutely should. The more interesting question is: Which work should we automate, and which work should we deliberately preserve as a learning opportunity?

AI can already conduct initial research, structure a presentation, summarize meeting notes, prepare an analysis, or draft a client email. That saves time. But what happens when early-career employees stop doing those things themselves?

Someone who has never done deep research will find it harder to judge whether a source is credible. Someone who has never built an analysis from scratch will find it harder to spot weaknesses in an AI-generated one. And someone who has never struggled to find the right words will have less instinct for what strong communication actually looks like. This is where AI literacy and AI fluency become so important.

Being good at working with AI is not just about writing better prompts or knowing how to use the latest tool. It is about understanding when AI is useful, questioning its output, recognizing errors, and knowing when human judgment needs to take over. And that still requires expertise, context, and experience. So perhaps we need to make a much clearer distinction when we think about skills development: Which capabilities can AI augment or take over, and which capabilities do people still need to develop themselves?

AI can make research dramatically faster. It cannot remove the need to evaluate sources critically. It can generate options. It cannot take responsibility for a business decision. It can summarize information. It cannot build human experience, empathy, or an understanding of organizational context in the same way a person can.

That is where I see one of the biggest risks. If we automate every task simply because we can, we may unintentionally automate part of the capability-building process as well. The challenge for organizations will be to think about efficiency, AI fluency, and human capability development together, rather than treating them as separate topics.

Conclusion: We Need New Ways to Build Experience

I don’t want AI to be understood as the threat in this conversation. Quite the opposite. If it removes repetitive work, speeds up processes, and gives people more time for higher-value tasks, that is a tremendous opportunity. We just need to make sure that, in removing the work, we do not also remove the experiences through which people develop expertise.

For a long time, much of that development happened naturally. Junior employees watched more experienced colleagues, worked through tasks alongside them, received feedback, and gradually took on more responsibility. Mentoring, apprenticeship-style learning, and interaction across experience levels were often by-products of everyday work.

If AI changes the way that everyday work happens, we may no longer be able to rely on those learning opportunities appearing on their own. We may have to build them deliberately. And for me, that leaves one big question: If AI takes over some of the work people have traditionally learned from, how do we make sure we are still developing the experts we will need ten years from now?

Sources:

Sondergeld, V: Berufseinstieg: Stellenangebot sinkt, Ausbildungssystem bietet einen Anker [Entering the workforce: Job openings are declining, while Germany’s vocational training system provides some stability]. Indeed Hiring Lab Germany. Indeed Hiring Lab – Berufseinstieg: Stellenangebot sinkt

World Economic Forum: Artificial intelligence and the future of entry-level work: A framework for safeguarding and reinventing early career pathways. World Economic Forum. WEF – Artificial Intelligence and the Future of Entry-Level Work

Mystery Minds: Building AI skills in organizations: Mentoring & peer learning. Mystery Minds. Mystery Minds – Building AI Skills Through Mentoring & Peer Learning

Mystery Minds: HR Trend 2026: Human Skills and AI Fluency. Mystery Minds. Mystery Minds – Human Skills & AI Fluency