Will AI replace juniors? The false debate that’s only the tip of the iceberg

I was discussing AI the other day with someone who was enthusiastic about the subject, to say the least. What he had to say was crystal clear and logical: “With AI, I won’t be needing junior developers any more, they’ll waste less of our time, and the seniors won’t have to spend as much time supervising them, so it’s all to the good.

My answer was obvious: “How do you expect to have senior staff if there are no juniors?

Who doesn’t dream of having only experienced, autonomous employees?

But unless you believe in a spontaneous generation of employees coming out of their mother’s womb with a Master’s degree and 10 years’ experience, it doesn’t work. A senior employee is someone who has first learned the basics, then learned the subjects required for his or her job, then how to apply them and acquire the necessary behavioral skills. This takes at least 30 years.

Even if we assume that everything can be learned online, there’ s still the stage of putting it into practice in the field. When you’ve learned something, you know it, but you don’t know how to do it, and that’s what your first years as an employee. And not just the first few, either: the world in which our original skills protected us for an entire career is long gone, and we’ll spend our lives not just learning, but learning how to put them into practice.

We’ll never have people who know and know how to do if we don’t have people who know without knowing how to do, and before that, people who know nothing. Or maybe you’re lost in the Baby Boss cartoon, in which case you have a different kind of problem.

The important thing is to know which jobs they are, and whether too radical a move will not result in a boomerang effect with, one day, a prejudicial scarcity of experienced employees.

Given that the subject was also addressed in an episode of Carlos Diaz’s excellent podcast, Silicon Carne (Intelligence Artificielle, le Grand Basculement !), which I absolutely recommend, it prompted me to put this topic back at the top of the pile of articles to be written one day.

AI and software development

Today, AI is already used on a massive scale to generate code (at Google, more than a quarter of the code produced is now generated by AI and then validated by engineers).

Generally speaking, it’s easy to see how AI could easily generate code for standard functions and simple applications, automate quality testing and intervene in debugging.

Of course, this will also lighten the workload for senior staff , but it’s easy to see that junior staff will be more affected.

While we are told that AI will be able to automate 60-70% of tasks (including in the software development sector) (The organization of the future: Enabled by gen AI, driven by people) , we need to be more nuanced.

Returning to development, and even if technology is improving day by day, it’s clear that AI’s contribution is reaching a ceiling for complex tasks : (Unleashing developer productivity with generative AI)

“Yet, while a massive surge in productivity is possible, our research finds time savings can vary significantly based on task complexity and developer experience. Time savings shrank to less than 10 percent on tasks that developers deemed high in complexity due to, for example, their lack of familiarity with a necessary programming framework. A similar result was seen among developers with less than a year of experience; in some cases, tasks took junior developers 7 to 10 percent longer with the tools than without them.”

Duly noted, but I don’t think it will last long.

And to come back to Google’s figures, we need to look at what they don’t tell us: isn’t a junior helped by AI progressing faster and perhaps doing a better job than a junior without AI, or even than some seniors? We’ve known for a long time, and since the Deep Blue era, that the strongest force is not the human or the machine, but the human who uses the machine (The rise of human-computer cooperation), so I’m taking the answer for granted: human-machine cooperation is the future (People and robots : what work relationships ?)… at least for a while.

So we wouldn’t necessarily be talking about the disappearance of juniors, but a kind of empowerment.

But let’s keep to the original question: the question is not whether some of the tasks performed by juniors will be automated (the answer is yes), but to what extent these juniors will still be needed, and to what extent not having been confronted with these tasks will make them less good seniors.

Does the automation of simple tasks destroy skills?

Will not having to code simple things one day prevent a developer from coding complex things?

Will the fact of never having to debug prevent him in the future from producing good code on more complicated functions and properly analyzing the causes of a malfunction?

I don’t know, I’m not a developer, but I’ll let you answer that.

In the consulting business it is said that around 30% of tasks in the sector are likely to be automated thanks to AI, particularly affecting junior positions in charge of data analysis and report preparation.

I don’t think that my years as a junior consultant spent churning out slides have given me a skill that I would miss today. On the other hand, knowing how to look for and swallow tons of data and information to make these slides does! I’m often told that I have a structured way of thinking and that I know how to look in the right direction. Some people call it instinct, others intuition, but for me it’s neither one nor the other. It’s just years spent in the shadows, doing menial work that at some point structured my thinking and indirectly gave me a certain body of skills. Some people see it as talent, but I only see it as the result of hard work, what we call experience.

Will the automation of data entry for administrative and legal professions prevent the people in question from becoming better professionals later on? No: these are tasks with no added value and jobs that only exist because we didn’t yet have the technology to automate them, and where the possibilities for professional advancement are limited.

Will automated CV sorting prevent young HR people from acquiring certain skills they’ll be lacking later on? The subject of the dehumanization of recruitment will be dealt with in its own time, but as far as the subject of skills is concerned, I’m not far from thinking that it will.

We could still go through a ton of professions, but things are far from being as black and white as one might think.

On the one hand, you have tasks with no added value that are only carried out by humans because we don’t yet have the technology to do them, and others that I still think are formative and will be missed by the accomplished professional that the junior aspires to become.

The risk of over-reliance on technology

We might indeed say that there will be no shortage of these skills, since AI will do the work anyway. But there are two limits to this line of reasoning.

The first concerns complex tasks , for which AI is still showing its limitations, as we have seen with developers. How can you perform a complex task if you haven’t learned to do the same tasks in a simpler version? How can a recruiter select a profile for a highly specialized need if he hasn’t learned to select simpler profiles?

And then there’s the risk of becoming dependent on technology, and finding yourself helpless the day it’s unavailable or breaks down. In everyday life, we use Excel or the calculator on our iPhone, but I challenge anyone to tell me that knowing how to do mental arithmetic is totally useless! Imagine a world where no-one could perform a basic operation in their head, it’s more worrying than anything else (although in France we’re heading straight for that, ditto for written expression).

Remember the blackout that blocked part of the world’s air traffic this summer? Apply this to almost any business and any task of low or medium complexity, and I think you can trigger a global economic crisis with a one-day blackout.

After the juniors, the seniors

But to ask the question of the loss of skills among future seniors without asking the question of their future usefulness, as we do for juniors, would be to answer a question while ignoring the globality of the problem of which it is a part.

Indeed, when I pointed out the limits of AI, I took the precaution of saying “today” and “for the moment”. Because tomorrow and the day after tomorrow will be very different.

Tomorrow is agentic AI. What are we talking about? Agentic AI makes decisions on its own, pursues goals and performs actions without continuous human intervention. It can also proactively plan actions to achieve its goals. It is capable of perceiving its environment, making decisions based on these perceptions and reacting accordingly.

And don’t be fooled into thinking that this is the future, because agentic AI is already starting to work in certain areas and weaving its web (Agentic AI and the future of process management).

After execution tasks, it’s the supervisory and decision-making tasks that will be impacted, those for which the famous experience that juniors need to acquire is necessary.

But agentic AI has its limits, not least because it is limited to a particular domain and specific tasks. Humans will therefore retain their superiority when it comes to general-purpose use cases, where the ability to adapt to any situation in any field and deal with the unexpected is required, and we have seen that AI is currently struggling in such contexts (AI challenges human CEOs in a simulation and gets fired).

However, it would be a mistake to tell ourselves that we’ve been saved and go about our business as if nothing had happened. After all, it only means one thing: we still have a little time to adapt to what’s to come. A little, but not too much.

What follows is AGI (Artificial General Intelligence ), a form of artificial intelligence capable of understanding, learning and adapting to any human intellectual task. It would be capable of solving problems in any field, learning new skills autonomously and transferring knowledge from one field to another.

This brings us closer to human intelligence and the ultimate goal of AI, and makes us realize that the loss of opportunity for junior staff to acquire skills is only the first step in a logic that leads us to the real problem:the obsolescence of the human being as a production tool.

And while we must always be wary of promises when it comes to technology, we are told that AGI will be a reality as early as 2027. Some even go so far as to say that it’s already here, but we can’t see it.

Society burns and we look the other way

As sometimes happens, I started this post with an idea in mind, and as I wrote and documented it, my thoughts were refined and led me into territory I hadn’t initially anticipated.

I had approached the subject from the angle of skills acquisition, which is a real subject, but one that I have to realizeis a short-term one. Solving the problem of skills, provided we can find a way of doing so, will not prevent us from hitting the wall, nor will it postpone the moment of impact.

As Carlos Diaz says, we’re at a tipping point. Listening to him and reading him (2026, L’Année où Tout Devient Obsolète?) made me realize something I’d been looking at but hadn’t seen (perhaps because I read too many foreign news sources): the vacuity of debate in France on the subject.

In France, we have the double talent of having brilliant thinkers and of missing the essential turning points (Anticipating the French way: when anticipating means reinforcing the status quo): we have a very good track record with the Théry report, which buried the Internet in favor of the Minitel, and the Riboud report (1987), which called for preparing its workers for the new challenges posed by emerging technologies. Here we are in 2024, and the issue of skills has never been dealt with proactively, but rather on a piecemeal fire-fighting basis, and tomorrow we’ll no longer be facing a skills problem, but a crisis of model from an economic and social point of view.

Are we surprised by the sudden arrival of AI? We’d have to be blind: as early as the early 2010s, we had already seen the beginnings of this essential reflection , with AI as a backdrop in the form of IBM Watson (Employement : a (lost ?) war against the machine ?). The thinking was there, but it didn’t take root in people’s minds due to a lack of urgency.

In my opinion, we won’t win the employment battle, so we need to reinvent our economic and social model , not tomorrow but today. Because if you think it’s enough just to adapt the current model, you’re in for a big disappointment (Towards a golden age of welfare and precariousness?)

I think it’s a mistake to treat as an intellectual or even technical subject what is an economic and social issue of the utmost importance, and one that threatens to devastate the social models we know. Needless to say, ethics will have to be at the heart of the debate, and that’s not what’s stifling the major players in the sector.

After that, it’s a question of political courage, but I’m afraid that our leaders are only remotely interested in the subject, without realizing the magnitude of what’s at stake.

Or do we think that technology itself, by virtue of its superiority, will generate the answers to the situations it makes possible? (To solve anything, click here) I doubt it very much..

Bottom line

I don’t think AI will totally destroy opportunities for juniors to enter the workforce: we’ll have fewer juniors, it’s true, but empowered juniors.

But we’ll need fewer seniors to supervise them.

And then AI will partly replace seniors.

Then it will be able to do almost everything a human can do, and we’ll realize that having approached the subject from the angle of skills and employment, we’ve missed the essential point of what’s going to happen to us.

It’s hard for me to imagine what this will look like and what it will involve, because we’re talking about something radical, something never seen before. But not a black swan because that’s predictable, it’s just the scale that I think is elusive to our current mental models.

Image:  scared employees by New Africa via Shutterstock.