Nobody wants to hire juniors anymore, and we need to do something about it.

If you’ve read the news lately, you’ve probably seen all the articles about how fresh grads are finding it hard to find jobs.

Job hunting without real working experience has always been undoubtedly tough, but this time it feels more systematic and more widespread across industries — tech, finance, and pretty much every knowledge-work industry in Singapore. Many companies and teams have quietly stopped hiring juniors, and I’ve seen this materialise in the teams within my organisation as well.

And for those of us in charge of hiring, let’s not pretend this is some great mystery.

Disclaimer: there may be (most definitely) bias towards my experience working in tech and government, but I believe my diagnosis and prescriptions are generalisable.

The Labour Equation Changed

There’s an uncomfortable calculus facing every hiring manager, whether they admit it or not: one senior engineer with AI tools can now do what used to take a senior plus two/three juniors. In many ways, the output gap has collapsed, but the training cost hasn’t.

Hiring a junior used to be seen as an investment - you’d spend six months getting them up to speed, pair them with senior mentors, accept lower output for a while, and eventually they’d become productive and valuable members of the team. This investment used to be even more valuable in the past where company loyalty was actually a thing.

Now? A senior with Claude Code and a good set of internal tools doesn’t need the headcount. The work just gets done at 10-100x efficiency. So when headcount planning comes around, the junior requirement is often to first one to get cut in hiring manager’s mind.

Add to that Singapore’s reality — where attrition in sectors like tech can be quite high, the job market (for experienced hires) is particularly competitive, and the manager who spent six months mentoring a junior only to watch them leave for a 20% bump at the company next door is not eager to do it again.

Everyone wants someone who can “hit the ground running.” But strangely enough, nobody’s building the ground. Or at least no one has the foresight to do so.

Why should this worry everyone? It creates a vicious cycle, because no junior hiring today means no mid-level pipeline in three years, which means no senior/lead pipeline in five years. We’re harvesting our roots and will be left wondering why nothing grows.

The Reframe

AI is making almost everyone (or at least those in the knowledge economy) more productive. But I suspect that we’re drawing the wrong conclusion from it regarding the labour market.

For seniors, the conversation should not be “great, now I can do the work of three people and coast.” It should be “the grunt work is automated, so what harder problems should I now be tackling?” If AI handles the boilerplate, the code generation, the first-pass, then the bar for what a senior delivers should go way up, not stay the same. Your AI tools are the floor now, not the ceiling.

So upskilling isn’t just a junior problem. If you’re a senior doing the same work you did in 2023 but with better tools, I’m sorry to tell you but you’re actually underdelivering, even if your company is not telling you so. It’s somewhat consistent with Jensen Huang’s recent comment that “he would be ‘deeply alarmed’ if his $500,000 engineer did not consume at least $250,000 of tokens”,

For juniors, the bar should shift too, but in a different way. A fresh grad armed with AI should be able to deliver what a mid-level engineer delivered three years ago in terms of raw technical output (and the caveat here is that we’re talking about raw technical output, nothing more). In essence, the execution gap is shrinking fast and this is a net good thing.

The way I see it, the real gap that remains is the one that no AI can automate away (at least for now). It’s things like product sense & intuition, stakeholder management, navigating ambiguity, best practices and processes, getting leadership buy-in, knowing when to push back, figuring out how to break up tasks, identify dependencies, manage risks, optimise for global maxima, the list of intangibles goes on, you get what I mean.

These are experiential skills. You cannot learn them from a tutorial, a course, or even from an all-knowing agentic assistant. You learn them by doing them, mostly badly at first, in a real workplace, with real stakes, iteratively until you develop the instincts and muscles. And you can only get those if someone gives you a seat at the table. That’s what my own experience in the field has taught me.

To me, that’s the crux of the whole problem with not hiring juniors. The skills gap is no longer primarily technical but experiential. And you can only close an experience gap by giving people the experience. (duh)

Which brings me to two rather radical proposals.

The Radical Moves

Radical Move #1: Force the Junior Quota

Mandate that companies hire a minimum percentage of juniors.

It’s actually not THAT crazy of an idea, especially in Singapore. We already mandate many things, especially in areas where the paternalistic state feels that the individuals/groups do not act in their own best (long-run) interests, or fail to internalise external societal interests. Even for the labour market, we have the Fair Consideration Framework - say what you want about it’s actual efficacy but I think it’s been rolled out in the right spirit. We have National Service, which is literally the government saying “you will spend two years doing something that benefits the collective, whether you want to or not.” My point is, we’re not a country that’s shy to lay down mandates when the collective good demands it, and when we have the foresight to see its worth.

Here’s how a junior hiring quota could work:

• Tiered by company size/characteristics. Companies above a certain headcount must maintain a minimum percentage of their workforce as people with under two years of professional experience.
• Adjusted by industry. Tech, finance, healthcare, manufacturing — they all have different realities, and the junior quota percentages should reflect that.
• Tied to existing incentive structures. Singapore loves grants, but I don’t think we employ them in the most efficient manner. We’ve got SkillsFuture, we’ve got IMDA/enterprise programmes, we have PWCS. If you ask me, some of these are not the best use of taxpayer money - for example, most people have no idea what’s a good use of their SkillsFuture credits. Just ask people around you what they spent their credits doing, if they even bothered at all. Wine course? Come on. Anyway, my idea is to tie the junior quota to enhanced grant access. Make it worth their while for the companies, and then make it mandatory. In an oversimplified way, think of it just like the Fair Consideration Framework, but for experience level. You already have to demonstrate fair consideration for local candidates, so just extend that logic to experience level.

I’d say this isn’t crazy because the skills that matter most in the long run, the stakeholder navigation, the product instinct, the organisational pattern recognition (everything other than the actual coding/writing) literally cannot be taught outside of a workplace. No amount of education reform solves this, and you have to get the fresh grad into the room.

Most companies won’t do this voluntarily because the incentives are misaligned. Every individual company benefits from letting other companies train juniors and then poaching the mid-levels. In some ways, it’s a classic tragedy of the commons on the labour market, we just have not realised it yet. And to my Economics folks, what’s the standard fix for tragedy of the commons? Good old regulation. It’s not an ideological left-wing prescription, but just basic economics of incentives and labour market failures.

The pushback on this proposal is somewhat expected, especially because you might argue that this would hurt economic competitiveness. But so does having zero talent pipeline in five years. If you think juniors slow your teams down, it’s only because you define their role the way we did in 2020. A junior with AI tools today is not the same as a junior five years ago. And if you are the free market economist who wants to let the market sort it out... well the market is sorting it out, in a terrible way. It’s sorting fresh grads into unemployment and companies into a slow-motion staffing crisis. The market has a tendency to optimise for short-term efficiency (because its agents are equally myopic), not long-term ecosystem health. So that’s what bold policymaking is needed for, more now than ever.

Radical Move #2: Overhaul How We Educate People

The junior quota solves the demand side of the labour equation. But we also need to fix the supply side, because right now, the education system is producing graduates trained for a world that no longer exists.

And no, the fix is not slapping an “AI module” onto an otherwise unchanged curriculum and calling it education transformation. To me that’s putting a paint job on a crumbling building.

Higher education needs to change drastically. I’m not saying throw out the fundamentals, because you still need to understand data structures, statistics, accounting principles, whatever your field demands. But restructure curriculum so that we focus less on the parts that AI can generate, retrieve and synthesise instantly, and instead have students focus on everything around it, understanding real-world systems, architectural concepts, planning & critiquing technical design documentation (not even the writing because AI can write out based on your plan), rather than memorise syntax for an exam or solve questions that look like the next Leetcode interview. To be honest, that’s something we should have fixed even before the age of accelerated AI but the 2020s just made everything more drastic.

The key is to accelerate students’ learning into the things that require real judgement - for engineers that’s stuff like system design, architecture decisions, knowing which framework to apply when and why, how to make decisions with incomplete information, how to communicate a technical concept to a non-technical stakeholder. Basically, anything that differentiates a useful employee from someone who can pass an exam. Because the harsh truth is that AI can pass most of your exams better than most of your students already. So if your curriculum is optimised for exam performance, you’re training people to lose to a machine. Again, these are known problems with the education system, but AI just brought the future closer to us.

Project work needs to be real, instead of capstone projects that end up in a drawer. If I were the Minister of Education, I’d upheave the curriculum to make at least 1/3 of education vocational and experiential - actual collaboration with actual companies on actual problems. Not internships where you make do a side project that was never meant to deliver anything, but structured partnerships where students work on production-adjacent problems and get feedback from people who ship things. In fact, many undergraduates are taking things in to their own hands by going for so many internships instead of focussing on their studies - they know for a fact that the grades don’t matter as much anymore.

And most critically, let’s push for conversion funnels. Look at how finance and law firms do it, at least when I was in the UK: summer programmes with real conversion rates into full-time roles. Some tech firms do this well too, but I think more firms can adopt this model, even beyond tech. The model exists, we just haven’t made it systematic across industries. In a nutshell, I believe that the apprenticeship shouldn’t be a box-ticking exercise for the university, but a meaningful first stage of a hiring pipeline.

AI fluency must be a foundational skill, not just using AI tools but understanding their strengths, failure modes, ways of evaluation, and practical integration modes into realistic workflows. This should sit alongside English and Maths as a core competency, and I see it as a non-negotiable skillset. And there should be universal, baseline access to AI tools for every student, not gated behind which school you go to or whether you can afford the subscription. If we’re serious about this being foundational, treat it like a necessity, maybe even a right, not a luxury.

Perhaps we should start even earlier. Primary school kids should be working with AI tools, secondary kids should be shipping prototypes. Instead of fighting against the tide of AI adoption (which you do see with teachers giving students F grades for using AI), think instead of how you as an educator can change the way your students learn and prepare them for the real world.

You Need Both, or Neither Works

The ideal state has a concrete vision: a fresh grad in 2030 should be able to walk into a company and operate at what we’d call a “senior level” today in terms of raw output and technical ownership. The experience gap, the soft skills, the org navigation, the product instinct: that’s what the junior quota is for.

That’s why these two proposals are not independent, they have to happen together to make it work for the kids going through our education system today.

Education reform without hiring quotas means you’ve trained highly capable graduates that nobody will hire, because the short-term incentives will still point away from junior hiring.

Hiring quotas without education reform means you’ve forced companies to hire people who still can’t deliver enough value to justify their seat. You’ve mandated cost without enabling capability, so companies will resent it, juniors will struggle with progression, and the policy will fail to realise it’s potential.

Thus, you really need both radical measures to make it work. Better-prepared graduates who can hit a higher bar on day one, and a system that guarantees they exercise the muscles they need to develop the skills that can only be learned on the job.

What I fear — and what I’m already seeing — is that we settle for the superficial version of action instead. You know the playbook: announce a new AI standing committee. Set up a task force. Fund a few AI startups and call it an “ecosystem strategy.” Create a digital academy that churns out certificates nobody takes seriously. Throw more SkillsFuture credits at people and hope they choose something useful (spoiler: they won’t - refer again to the wine course). Commission some R&D that produces papers but no pipeline. None of this moves the needle on the actual structural problem, which is that companies aren’t hiring juniors and universities aren’t producing graduates who can bridge the gap. The government has to stop mistaking activity for progress. Standing committees don’t create jobs, and task forces don’t teach product sense. And no amount of grant money for AI startups or courses solves the fact that a capable, ambitious 22-year-old with a degree can’t get their foot in the door.

So What Now?

If you’re a hiring manager / senior leadership: We have to firstly acknowledge that the junior hiring problem IS your problem. Otherwise, you’re just deferring the cost, and the interest rate on that deferral will hit your organisation one day. Start thinking about what a junior role looks like in an AI-augmented world — it’s not the same as it was, and you need to see that as a good thing.

If you’re a fresh grad: The bar is higher, but so is your ceiling, and learn to use AI as a multiplier in your personal and professional life. Build things, ship things, don’t over-plan, don’t wait for permission or for someone to give you a structured learning path. The most hireable juniors in 2026 are the ones who’ve already demonstrated they can deliver, with or without a job title.

If you’re a senior: Your seniority is no longer a moat. If a junior with AI tools can do what you did last year, you need to be doing harder things this year. It’s an opportunity to flex the higher-level thinking, product-wide muscles that makes you a more effective senior wielding the AI multiplier.

And finally to my friends in policy or education: My key message is that the system is producing graduates for 2015, but we need graduates for 2030. You might think that the pace of AI acceleration is fast today, but you and I have no idea what’s gonna hit us in 2030. Whatever it is, it’s gonna be beyond our current imagination. For those practitioners in the AI field, you know the gap is already here and it’s getting wider every year. Another task force made up of our most senior politicians isn’t going to fix this. Another round of SkillsFuture top-ups isn’t going to fix this.

We need structural reform, of curricula, of hiring incentives, of the entire pipeline from school to workplace, and we need it urgently. The pipeline is beginning to run dry while we’re still forming committees to catch up with reality.

When the problem in the labour market is deeply structural, the response needs to match. Actual, uncomfortable, structural change surrounding the way we hire and educate is needed.