Losers in the Job Market? New Graduates vs. AI
“I had no idea you had to have been a horse first to become a good jockey.” With this legendary line, soccer coach Arrigo Sacchi once silenced his critics, who accused him of lacking professional experience as a player. His analogy perfectly describes the dilemma that is sweeping the entire world of work today, in 2026. Because we’re facing a massive problem: let’s call it the junior skills crisis.
The Iron Law of Hard Work
Until now, career development has followed an unwritten rule: You start at the bottom. You do the tedious, grind-like work. As a junior auditor, you reconcile account balances; as a junior developer, you spend hours searching for semicolons in the code; as a junior copywriter, you write SEO copy about vacuum cleaner bags.
The prevailing view in labor market research is that only those who learn this routine work from the ground up develop the necessary depth to make well-informed decisions later on as senior professionals. The hard grind was considered the only path to expertise.
That has undoubtedly been the case so far, but only because it made economic sense to have lower-cost employees perform tasks that needed to be done anyway. It was a pragmatic stopgap measure, not a well-thought-out educational concept. And this approach is currently undergoing a radical shift.
The Break-In at the Base
A recent Lünendonk survey shows that two-thirds of accounting firms plan to cut entry-level positions. Stepstone reports that job postings for entry-level candidates have plummeted. Why? Because AI handles these routine tasks faster, more cheaply, and often with fewer errors.
This is extremely tempting for companies. But researchers warn: If you don't hire junior employees today, you'll be short on senior employees in five years.
This leads to two hypotheses that we urgently need to discuss.
Hypothesis 1: The Undiscovered Path to Training
We haven’t even found the true path to “verification expertise” yet. The real core competency of the future isn’t execution, but leadership and verification. Today, a senior professional must be able to critically evaluate the output of an AI, identify errors, and draw strategic conclusions.
Dr. Fabian Stephany, a renowned labor market researcher at the University of Oxford, suggests a similar—albeit more cautious—approach in a recent Handelsblatt article: Companies should shorten the training period for junior employees, introduce them to specialized knowledge more quickly, and assign them project responsibilities earlier.
Here at Leaders of AI, we’re taking it a step further. We don’t just believe that the time required can be shortened. We believe we haven’t even found the actual training path for verification expertise yet, because we’ve never had to build it.
Jockeys from Day One
For years, it was generally assumed that you had to dig through code or spreadsheets yourself to learn this work. That's a misconception.
At SAP, we’re already seeing this: Junior developers like Stefanie Lanz no longer get bogged down for hours in manual debugging. Instead, they use AI to analyze the code and discuss the software architecture and risks with senior developers right from the start. They learn critical thinking and strategic analysis—in other words, true verification expertise—in record time because they simply skip the monotonous grunt work.
Hypothesis 2: The ability to download
What if competence were less an experience one goes through and more a process that can be observed and recorded?
That’s exactly what’s happening right now from a technical standpoint. AI agents can watch an expert at work, log every click, every query, and every decision, and use that data to build a repeatable workflow. Anthropic has implemented this with the “Record a Skill” feature for Claude: A senior expert demonstrates once how to solve a task. The agent then performs the task independently.
(💡 You can find more about “Record a Skill” in our blog post“How AI Helps You Preserve Baby Boomers’ Knowledge”.)
All of this is not yet a blueprint of the human brain. It is the first serious attempt to translate implicit knowledge—that is, what an expert simply knows without being able to explain it—into an explicit, transferable format.
For Leaders of AI, this means: We’re teaching junior staff to use Excel spreadsheets differently today. We’re equipping them with AI agents that already contain this documented operational knowledge.
The junior thus starts out at the verification expert level. His or her job is to check the results for plausibility and draw the correct strategic conclusions.
The Conductors of Algorithms
This is precisely the sweet spot that we systematically teach in our training programs (MBAI, AI Integration Expert, AI Survival Program): It’s no longer just about typing blunt commands into a chat window. The real superpower of today’s young professionals lies in using AI models with a radically critical eye while simultaneously unlocking their own, purely human strengths.
This leadership skill can now even be measured. A recent study by the Harvard Kennedy School shows that those who successfully lead a team of AI agents through a complex task—by asking more questions, maintaining dialogue, and thinking strategically—also lead teams of humans more successfully. The correlation between the two skills is 0.81, an exceptionally high value for behavioral studies in the social sciences, where correlations are often significantly lower.
This aligns with what we at Leaders of AI consistently observe in our training programs: Those who practice daily—leading AI agents precisely, asking the right questions, debunking hallucinations, and focusing on strategic judgment—are honing precisely the skills that matter most in “human teams” as well. Learning by doing— only with a team that never gets tired and never quits.
Anyone who learns this becomes a true conductor of algorithms. They won't be replaced by AI. They'll lead it.
The End of the Grind
The end of the backbreaking work is a relief, not a loss of quality. We need to stop complaining that young people can no longer sort receipts.
Who can still reliably multiply a three-digit number in their head these days? Hardly anyone, ever since calculators came along. (Editor’s note: One of the greatest inventions of all time.) Who can still find their way through an unfamiliar city without a GPS? Hardly anyone, ever since GPS came along.
Both of these skills have all but disappeared, and no one seriously misses them. The real question was never whether we should practice mental math or map reading. It was why. As soon as a machine can do it more reliably and quickly, the old skill becomes a museum piece, and the time freed up can be spent on something more valuable.
That’s exactly what’s happening right now with sorting receipts, debugging semicolons, and writing SEO copy about vacuum cleaner bags. When we redesign the transition into the workforce—moving away from routine tasks and toward early responsibility and critical thinking—we ’re building a completely new, better kind of expertise . One that uses AI as a tool rather than fearing it.
After all, at the end of the day, it's the jockey who wins the race, not the horse.
Sources
- Franziska Telser, Milena Merten, Bert Fröndhoff: “The Job Market: How AI Is Making It Increasingly Difficult to Start a Career,” Handelsblatt, 2026, handelsblatt.com/karriere/arbeitsmarkt-wie-ki-den-berufseinstieg-immer-schwieriger-macht/100238206.html.
- Fewer Entry-Level Jobs: Is Gen Z Facing a Career Crisis?, t3n, 2025, t3n.de/news/weniger-einstiegsjobs-steckt-die-gen-z-in-einer-beruflichen-krise-1702542.
- Anthropic (@claudeai): Record a Skill, product announcement, X, July 21, 2026 (see also: Claude's Record a Skill: Train Agents by Demonstration, digitalapplied.com/blog/claude-record-a-skill-train-agents-by-demonstration-2026).
- Ben Weidmann, Yixian Xu, David J. Deming: "Measuring Human Leadership Skills with AI Agents," Harvard Kennedy School / NBER Working Paper, 2025, hks.harvard.edu/publications/measuring-human-leadership-skills-ai-agents.
- Arrigo Sacchi's Best Quotes, Inside FIFA, 2016, inside.fifa.com/de/news/arrigo-sacchis-beste-spruche-2773467.
