More Than a Jobs Crisis — AI Is Severing Hong Kong’s Talent Pipeline

Secretary for Labour and Welfare Sun Yuk-han disclosed a set of figures in the Legislative Council earlier this month: full-time job vacancies suitable for university graduates fell from 80,000 in 2022 to 31,000 in 2025, a drop of more than 60%. Administrative vacancies have nearly evaporated, with a decline approaching 90%; information technology and programming positions have also shrunk by 80%. Secretary Sun stated plainly that the spread of artificial intelligence is a major reason.

These figures are striking. But what they reveal goes far beyond a jobs crisis — AI is changing the fundamental demands the labour market places on talent, yet Hong Kong’s entire system for nurturing talent still operates on the old logic.

1. Entry-Level Workers: First to Be Hit

The spread of AI has not fallen evenly. Research from Denmark and the United States shows that in heavily affected industries, the pay of senior employees has been largely unaffected; meanwhile, demand for entry-level positions centred on organising information, drafting documents, and basic coding has dropped sharply. The reason is straightforward: AI learns from massive records, and can handle any work that can be recorded and verbalised. The steep decline in administrative, clerical, and IT programming positions is precisely this logic playing out in Hong Kong.

Yet there is a deeper economic logic behind this shock. As AI continues to take over verbalisable work, the value of another kind of ability keeps rising — judgement and perception that can only be honed in real situations. Clinicians reading unvoiced anxieties from a patient’s expression, teachers noticing when a student has mentally drifted, social workers judging a family’s true condition during a home visit — these abilities are constrained by a person’s time and experience, and cannot be copied or scaled. The more restricted the supply, the higher the value. Book knowledge and procedural skills are relatively devalued; judgement forged in practice is relatively appreciated. Today’s seemingly unshakable senior professionals are, in fact, the beneficiaries of this shift.

2. Learning to Use AI Is Not Enough to Stay Competitive

Facing the AI wave, the government has not been idle. The Education Bureau’s “AI in Education” funding programme provides each public-sector school with HK$500,000 to promote the integration of AI in teaching; the eight funded universities will add around 30 related courses within three years; HKU and HKBU have already required undergraduates to take AI literacy courses; and the Employees Retraining Board has been restructured as “Upskill Hong Kong”, with AI application as one of its core components. The direction is right, and the effort is not small.

But these measures broadly fall into two categories: bringing AI into the classroom, or training people to use AI tools well. Both target the same question — how to make people and AI work better together.

What is truly being overlooked is another question: how to nurture those abilities that AI cannot replace — judgement forged in practice, sensitivity in dealing with people, and professional experience that can only be accumulated through first-hand experience.

This is not just a question of curriculum design, but of institutional incentives. Hong Kong’s higher education has long done the same thing: efficiently transmitting codifiable knowledge, and using standardised examinations to verify outcomes. In the past this made sense, because such knowledge was highly valued in the workplace. But AI is rewriting that equation. The UGC’s funding indicators measure research output, staff-student ratios, and graduate employment rates — none of them rewards mentorship-based learning, project-based teaching, or situational assessment. If the system does not change, behaviour will not change.

3. Entry-Level Jobs Disappear, and the Talent Pipeline Snaps With Them

Even if the education system successfully transforms, there is a deeper question that has yet to enter the policy horizon.

Entry-level positions have never been just “cheap labour”. Young lawyers build legal language intuition through repeatedly drafting contract memos; junior doctors cultivate clinical judgement through large volumes of patient consultations; management trainees learn how organisations work through the mundane daily business. This kind of learning is rooted in real action and responsibility for consequences — it cannot be acquired from textbooks, nor simulated by AI. The apprenticeship period is precisely where society cultivates “talent that AI cannot replace”.

As vacancies shrink, this channel narrows with them. AI has indeed greatly boosted productivity, but the surface gain obscures a more hidden loss: opportunities for young people to accumulate real-world experience are quietly diminishing.

The problem has structural roots. When a company trains newcomers, it is in effect cultivating professional talent for the whole of society; but once employees complete their training, they can move on, and the company cannot keep the returns to itself, so it naturally does not want to bear the cost alone. The arrival of AI has dramatically worsened this problem — when machines can handle most of the entry-level work, the incentive to hire newcomers drops sharply.

Even more intractable is the collective-action dilemma. Even if every company clearly understands that the industry needs a healthy training mechanism, none is willing to move first — the talent painstakingly cultivated can be poached by competitors at any time. The result is that everyone watches, and no one moves. The market cannot self-correct; external force is needed to break the deadlock.

Ten years from now, will Hong Kong still be able to cultivate the next generation of senior lawyers, doctors, and engineers? This question deserves far more serious treatment than it currently receives.

4. Expand Apprenticeships to Hold the Foundations of Talent Development

Hong Kong actually has successful precedents. The key to breaking the collective-action dilemma lies with regulators or professional bodies bringing universities, industry associations, and employers together, so that no single party has to bear the risk of going first alone.

In engineering, the Vocational Training Council and the Hong Kong Institution of Engineers jointly run the Engineering Graduate Training Scheme (EGTS), covering a wide range of engineering disciplines, providing companies with salary subsidies of up to 18 months, with about 300 places each year.

In finance, the Hong Kong Monetary Authority and the Private Wealth Management Association co-run an apprenticeship scheme, providing at least 8 weeks of comprehensive training each year for 45 students, with 17 major banks participating.

These programmes prove the model is workable. The problem is that the scale is too small.

The eight funded universities produce about 4,000 engineering and technology graduates each year, so the EGTS’s 300 places cover less than 10%. Across the territory, about 30,000 university graduates enter the labour market each year, but the total number of structured apprenticeship places across industries is extremely limited, dwarfed by the graduate numbers. These schemes were a nice-to-have in times of plentiful vacancies; today, with vacancies sharply reduced, their scale is far from enough to fill the gap.

There are two directions worth exploring. First, significantly expand existing programmes — increase EGTS places, raise subsidy levels, and push the HKMA model into more industries. Second, extend the framework to areas currently lacking apprenticeships — management consulting, technology, creative industries and so on, which absorb large numbers of humanities and business graduates, and which are also among the industries facing the most severe vacancy contractions.

There is in fact an existing institutional foundation. The Vocational Training Council (VTC) has 25 training committees covering major industries, with committees for banking and finance, accounting, and information technology already bringing together the HKMA, regulators, industry associations, and universities. The problem is that these committees have so far mainly played a curriculum-advisory role, and have never been activated as a mechanism to design and drive structured apprenticeship frameworks. The Government could authorise and fund the VTC training committees to take on this function — significantly expanding apprenticeship scale in industries already covered, and setting up new committees in those not yet covered, to fill the gaps.

This is not short-term employment relief, but a medium- to long-term investment in talent development. Hong Kong already has the institutional architecture to be activated; what is missing is the political will and resources.

5. Conclusion

Secretary Sun’s figures describe a visible problem: today’s graduates cannot find suitable jobs. But the cost goes well beyond that. The disappearance of vacancies is simultaneously severing the main channel through which society cultivates real-world talent — and that, precisely, is the scarcest and most valuable thing in the AI era. This loss is hidden, and will only become apparent in ten years’ time.

Promoting AI skills training is right, and should not stop. But if the policy discussion stops there, we will miss the deeper challenge: AI is changing not just which jobs exist, but what kind of talent society needs. Neither our education system nor our corporate training mechanisms are yet ready for this new reality.