The Missing Middle: The Hidden Cost of Not Hiring Juniors in the Age of AI

Picture a company in 2033. It has rebuilt most of its operations around AI agents. Customer service, reporting, content, first-level analysis: everything runs, and mostly well. Then someone on the board asks a simple question. Who is actually responsible for all of this? Who checks the agents, catches their mistakes and answers for the results? The company starts looking for experienced people to do that job, and realises there are very few of them. Every competitor is looking for the same profiles.

This scene is imagined. The decisions that could lead to it are not. They are being made right now, quietly, in hiring plans that look perfectly reasonable in a spreadsheet. In this piece I try to follow that chain of decisions over the next ten years. It is not a forecast. It is how I read what I see around me, and I may well be wrong on the timing.

2026: a door closing from the outside

If I had to describe the job market of 2026 in one sentence, I would say this: people who have a job keep it, and people who don't have one struggle to get in. Companies are not firing much. They are simply hiring less.

What strikes me is who pays for it. It is not the experienced people. It is the young ones. Graduates are finding it harder to land a first job than the generations before them, and it seems to hit hardest in exactly the kind of work AI is good at: first drafts, first analyses, first answers to customers.

And the way it happens matters. Nobody is being pushed out. The door is just not opening anymore. That is why so few people are alarmed. There is no wave of layoffs to point at. There are only the jobs that were never posted.

At the same time, more and more companies say openly that AI is the reason they need fewer people. We should be careful with that. Some of it is real, and some of it is a convenient way to present ordinary cost-cutting, because AI sounds better to investors than saving money. AI is not the only cause either. Interest rates, and the over-hiring that followed the pandemic, explain part of the slowdown. But none of this changes the direction. It only changes how fast things move.

2026 to 2031: from adding AI to rebuilding around it

The first years of AI in companies were mostly experiments. Lots of pilots, lots of demos, very little real impact on the numbers. From what I have seen, the problem was rarely the technology. It was that AI got plugged into old ways of working instead of changing them.

What I find striking is that hiring changed before the results showed up. Some leaders told their teams to prove AI could not do a job before asking for a new hire. That logic spreads very easily, because it costs nothing to the people already in place. It just stops adding people.

The next phase is a real redesign. Companies will stop bolting agents onto existing processes and start rebuilding the processes around what agents do well. Only then will they decide which humans they need, and where.

This redesign changes the shape of organisations. Fewer people at the base doing first versions of things. More people in the middle designing the workflows, reviewing what the agents produce and taking responsibility for the result. Judgment stops being something we hope senior people have. It becomes the core of the job.

Most long-term forecasts say AI will create more jobs than it destroys. I hope they are right, and on the whole they probably are. Others warn that a large part of entry-level office jobs could disappear within a few years. What I notice is that both camps agree on one thing: the entry level changes first and fastest. The real question is not whether the base of the pyramid shrinks. It is what happens to the people who would have climbed it.

2033: the missing middle

This is the part that worries me most, and the one I rarely see discussed.

Junior work was never only about output. The first version of the presentation, the analysis nobody else wanted to do, the customer tickets, the boring reconciliations. That is how we learned. Every task was a repetition, and enough repetitions turned a beginner into someone with judgment. Looking back at my own career, I learned most of what I know by doing exactly that kind of work.

AI now does those tasks, and does them well. So companies keep the output and quietly lose the training. A junior working with AI today can produce more than any junior before, while getting fewer of the repetitions that used to make people experts.

Now let's play it forward. The juniors who are not hired in 2026 would have been mid-level around 2029, and senior around 2033. They won't be there. I can't prove this, but I find it hard to argue with the logic: senior people are made from junior people. There is no other source.

And here is the paradox. An organisation built around AI agents needs more experienced judgment, not less. It needs people who can tell a good answer from a confident mistake, who understand the business well enough to design the workflow, and who can take responsibility when it fails. AI gives the most to the people who already know their field. It rewards expertise. It does not create it.

That is what I call the missing middle. Not a shortage of jobs, but a shortage of the experienced people needed to run the new ones, created by decisions that looked like efficiency at the time.

The Gulf is running on a different clock

Living in Dubai, I have to admit that this story looks very Western from here. The Gulf seems to be running on a different clock. The driver here is less about companies cutting costs and more about massive investment in AI infrastructure itself: data centres, computing power, national programmes.

Hiring has not followed a straight line either. The region had a strong start to the year, a difficult period during the regional conflict, and then signs of recovery. My impression is that hiring here moves with investment cycles and geopolitics much more than with the quiet replacement of junior roles I see described elsewhere.

So I keep asking myself a question I can't answer yet. Is the Gulf skipping this freeze because it is building the infrastructure itself? Or is it simply a couple of years behind the same curve? If it is the second, the region has a real advantage: it can see the missing middle coming, and design around it before it arrives.

2036: two possible endings

When I try to imagine the mid-2030s, I see two broad scenarios.

• Apprenticeship, reinvented. Companies deliberately create learning roles, where the goal is to build judgment rather than to produce. Juniors review what agents do, own small workflows from start to finish, and learn the business by being responsible for real results early. Training becomes a strategic investment, a bit like R&D today.
• The experience premium. Nobody coordinates. Every company makes the rational choice not to train, experienced people become rare, and their price goes up a lot. The gap between those who got in before the door closed and those who didn't becomes one of the big divides of the decade.

The first ending will not happen on its own. It asks companies to invest in something that pays off years later, and might even end up benefiting a competitor. That is exactly why it is worth deciding now.

What I would do as a leader

If you are in charge of hiring, the question is not whether to use AI. The question is what you do with your entry level once AI can do its tasks. Here is where I would start.

• Keep hiring juniors, on purpose. Treat it as an investment in future judgment, and size it deliberately, instead of making it the first thing to cut.
• Redesign junior roles around judgment. Give juniors the job of reviewing, correcting and owning what AI produces, rather than the tasks AI now does on its own.
• Measure learning, not only output. Look at how fast people become autonomous, not only at how much they produce with AI.
• Put juniors next to the people who own AI workflows. The people responsible for the agents are the natural teachers of the next generation.
• Name an owner for every AI workflow. Responsibility is where judgment is built. Without an owner, nobody learns anything when something goes wrong.

Where I might be wrong

I want to be honest about the limits of this reasoning. These are the things that would make me change my mind.

1. The promised gains never come. If most AI projects keep failing to pay off, companies will start hiring juniors again, and this freeze will look like a hype cycle.
2. AI is mostly an excuse. If AI is mainly a nice label for ordinary cost-cutting, this is an economic cycle, not a structural change.
3. The technology stalls. If AI stops progressing as fast, every phase stretches out and organisations get more time to adapt.
4. Governments react. Rising youth unemployment is exactly the kind of problem that triggers fast political answers, from hiring incentives to apprenticeship programmes.
5. The economy matters more than AI. Interest rates and the business cycle may explain more of the slowdown in junior hiring than technology does.

Each of these would change the speed. None of them changes what I believe at the core: expertise is built through practice, and practice is what the entry level used to give us. Whatever the timing, that is what we should protect.