The deepfake CEO: Why leadership teams may have a false sense of security
Effective defense against deepfakes depends on how CEOs shape behavior, communication, and decision-making under pressure, says Öykü Işık....
Talent
How to build judgment when AI does the work
If entry-level roles no longer train graduates through routine tasks, companies will need new ways to develop the next generation of leaders.
Rapid read:
- AI is removing routine entry-level tasks, reducing hiring and limiting traditional pathways for graduates to build experience.
- This creates a “judgment gap,” as early-career workers lose opportunities to develop tacit knowledge through practice.
- Companies must redesign roles, training, and leadership development to ensure future talent pipelines remain strong.
If artificial intelligence removes the bottom rung of the career ladder, how will graduates develop the judgment they need to progress? And how should companies redesign the roles that once built it, seeding their leadership pipelines?
Already, the pressure is showing up at the bottom of the labor market. A recent analysis by the Federal Reserve Bank of Dallas found employment in the most AI-exposed sectors has fallen 1% since ChatGPT’s launch in late 2022, even as the rest of the economy grew 2.5%. Research from Stanford University suggests the decline is concentrated almost entirely among workers under 25, driven less by layoffs than by a sharp drop in hiring.
In the UK, there are about 140 applications for every graduate vacancy, according to the Institute of Student Employers.
The judgment gap
Much of the work that once trained junior employees – from drafting documents to building models and compiling research – can now be handled by generative tools.
For lower-level staff, this shifts the focus from producing content to checking results and taking responsibility for the outcome. And that places a premium on judgment, which is something that is built through years of practice. It cannot just be installed like software.
In practice, judgment means knowing when you need to question an answer a machine gives you. Some work problems have clear answers; others are more complex. So, as AI takes on more of the former, much early-career work will shift towards the latter.
For example, a junior analyst might use AI to summarize a market for a client. The answer may look clear and convincing. But good judgment means checking whether the information is up to date, whether the sources are real, and whether it answers the client’s question. Poor judgment is accepting it because it sounds right. Good judgment is treating it as a first draft.
The distinction is familiar in economics: codified versus tacit knowledge. AI excels at the former: tasks that are procedural, rule-based, and easy to check. Tacit knowledge, by contrast, is built through experience, error, and feedback, and remains harder to replicate.
The same divide is visible in the data. Analysis by the Dallas Fed shows wages in the most AI-exposed occupations have risen 8.5% since 2022, above the national average. The gains accrue largely to experienced workers. Where entry-level and senior wages are similar, and tacit knowledge is less important, exposure is already putting downward pressure on pay.
The question for organizations is how that tacit knowledge is built if AI removes the entry-level work where it was once learned.
Companies will need new ways to develop judgment in their fresh recruits.
The pipeline problem
Companies will need new ways to develop judgment in their fresh recruits. Much of that learning came from doing the drafting itself, often without much formal training. So, if companies expect to need staff who can handle more complex work in three or four years, they need to build those skills now – deliberately.
One option is to bring in graduates to shadow more experienced staff, learning the parts of the job that AI cannot do, such as how decisions are made, how trade-offs are weighed, and how clients are advised. This, of course, has cost implications and forces organizations to explicitly budget the cost of “learning on the job.”
That comes at a time when some companies are cutting entry-level roles. Even so, they’re unlikely to vanish. Someone still needs to monitor and manage what the machines generate. That points to a different kind of role. Graduates may use AI for analysis, while managers focus more on teaching them how to interpret the results and make decisions.
In practice, this means less time producing drafts or compiling data, and more time interrogating what AI creates, looking for gaps, questioning assumptions, and tailoring it to the audience. Juniors may spend less time preparing a presentation, for instance, and more time in the room where it is reviewed, seeing what gets challenged and why.
This shift also changes the role of managers. In the past, they coached juniors by reviewing their work and showing their thinking through edits. That becomes harder when AI produces the first draft.
Managers now need to coach differently. Instead of correcting work, they need to explain how decisions are made and bring juniors into discussions earlier so they can see what gets challenged and why.
Without that shift, companies risk weakening their future leadership pipeline. Organizations that shrink their entry-level intake are running an experiment whose consequences may only become visible years later.
IBM plans to triple graduate hiring in 2026. The US tech company warned that reducing junior intake today risks a shortage of experienced managers within a few years.
No settled model
There is, as yet, no clear way to train and develop junior employees when AI is doing much of their work. Most firms can see the problem, but few have solved it, and it is too early to know what works. For now, companies are experimenting.
For example, IBM plans to triple graduate hiring in 2026. The US tech company warned that reducing junior intake today risks a shortage of experienced managers within a few years. The company has also redesigned roles so junior developers spend less time on routine coding and more time working with clients, across teams, and contributing to new products.
Dropbox, meanwhile, is also increasing its internship and graduate intake by 25%, with the cloud storage provider viewing younger workers’ familiarity with AI as an advantage.
Some best practices are emerging. In some firms, AI’s output is treated as a starting point, not the final answer. Juniors are expected to bring both the output and their view on it: what they changed, what they doubt, and what needs checking.
Some companies are also holding back from full automation. Not every task is handed to AI, because some work exists to build judgment rather than deliver a result. That may seem inefficient, but it is an investment in experience; the kind of knowledge that can only be built over time.
Junior staff are not just learning; they are also a source of insight. They use AI tools more than anyone else, and what they see (where outputs are wrong or incomplete) helps firms understand how much they can rely on them.
These early responses point to a broader choice. AI can remove the bottom rung of the career ladder, and for many firms, the short-term logic is compelling. But that rung was never just about output. It was where people learned to think and build judgment. Firms that cut entry-level roles without replacing that training ground risk a problem that will only show up years later, when today’s missing juniors fail to become tomorrow’s experienced managers.
The task now is to rebuild that training ground deliberately – before the window closes.
The degree model under pressure
The traditional model is to study first and learn on the job. But if entry-level roles no longer provide the same experience, that sequence breaks down.
Graduates may need to alternate between study and work to build skills that routine tasks once provided. That points to shorter cycles of study and work, and to degrees completed alongside employment.
AI training will also need to be embedded across all disciplines, because using and judging what machines generate will become part of almost every job.
Authors
Related
Want to survive your back to backs? Choose the right microbreak
Microbreaks reset energy and avoid depletion. But which type of break do you need? This simple, three-step framework will help you decide, and power through your New Year back to backs. ...
Tough at the top: Why CEOs are moving beyond invincibility
As pressure on chief executives grows, long-held assumptions about toughness are being challenged. IMD’s Nele Dael explains why it is important for those at the top of organizations to be open about...
The human change challenge: How Sanofi gets workforce buy-in for AI
As transformation accelerates, HR leaders should keep their people up to speed, explains Sanofi’s Raj Verma....
Learn Brain Circuits
Join us for daily exercises focusing on issues from team building to developing an actionable sustainability plan to personal development. Go on - they only take five minutes.
Explore Leadership
What makes a great leader? Do you need charisma? How do you inspire your team? Our experts offer actionable insights through first-person narratives, behind-the-scenes interviews and The Help Desk.
Join Membership
Log in here to join in the conversation with the I by IMD community. Your subscription grants you access to the quarterly magazine plus daily articles, videos, podcasts and learning exercises.