How recent graduates entering the job market can outperform AI

Artificial intelligence (AI) is eroding the very capabilities young professionals need to remain valuable in a machine-automated era. That is the real challenge facing graduates entering the workforce. It is not simply a case of AI destroying jobs – though it is – but that it is undermining our cognitive and interpersonal skills. The numbers are starting to confirm what we suspected. Hong Kong graduates in 2025 found 55 per cent fewer job opportunities than the year before. Over 12 per cent of those aged 20 to 24 were unemployed in December, the second-highest figure on record. Given a roughly 40-fold reduction in the cost of synthetic cognition (AI) per unit of output, a 2023 prediction that 25 per cent of the city’s workforce would need to change careers by 2028 now looks conservative. Many of Hong Kong’s thousands of legal jobs are threatened by products like Anthropic’s recent plug-in for legal work. Other local applications are bound to come. Hiring managers make decisions based on economics, not social good. They need to be convinced that a young human worker is a better investment than a machine. Many employers are reluctant to bet on people with no practical experience and no domain knowledge. Management may prefer to upskill experienced staff to use AI to do more of what they already do. A perceived preference for job-hopping among younger workers only compounds the disincentive. This conforms to our experience as a small company paying high Hong Kong rents. The “productive struggle” of thinking through hard problems, failing, retrying and eventually making breakthroughs has always been at the core of genuine learning. That this struggle is increasingly bypassed is what keeps educators and developmental psychologists awake at night. It’s also what troubles managers when junior staff expect work to be easy. The friction of navigating daily life, of developing street smarts, is being removed for AI natives. Those of us who remember life before smartphones know that screen time rewires the brain for immediacy and surface-level skimming. Cognitive offloading, the outsourcing of energy-intensive intellectual effort to technology, is a natural human reflex. But it leads to what researchers call cognitive debt: the gradual atrophy of memory, creativity and critical thinking. The temptation grows when AI assistants live in every app. When chatbots can produce 20-page reports in minutes, they foster impatience, shallow understanding and an unwillingness to do extra work. One technologist at The Neural Information Processing Systems Conference recently suggested that we are moving from “solving intelligence” to “using intelligence to solve everything else”, including the hardest problems in maths, science, engineering and medicine. Predicting a single protein’s structure used to take years. AlphaFold is swiftly working on solving virtually all known protein structures. Reasoning models are working their way through the unsolved mathematical list of conjectures known as the “Erdos problems”. If machines can do that, what happens to a young scientist’s, lawyer’s or analyst’s willingness to endure years of productive struggle? The prisoner’s dilemma at the company level ensures the pace will not slow down. The pressure to monetise enormous AI investments, combined with geopolitical competition, means the assault of the machines is unlikely to be curtailed. Everyone, including policymakers, business leaders, educators and young professionals themselves, must recalibrate. Time is short, and a generation of unemployed graduates is waiting. The government must act to derail a train bound to crash into aspiring young graduates. One way to start is to urgently rethink and retool existing subsidy frameworks. Hong Kong’s Research Talent Hub provides salary allowances for R&D roles. Rather than allowing firms to merely offset the cost of low-level technical headcount, the subsidy should be conditional on evidence of “cognitive apprenticeships”, encouraging a shift away from a focus on cheap labour towards a deliberate investment in building capability. Bolder still, regulators should introduce cognitive capability standards in professional licensing and accreditation. If you want to practise law or accounting, you must show you can reason through a complex problem without an algorithm. This would create artificial market demand for human skills at risk of atrophy. Employers must redesign junior roles. Just as pilots must learn to fly manually before relying on autopilot, companies can structure early-career positions so that AI helps rather than generates, reducing the extra work to parse “AI slop”. New hires can be required to reason through problems and use AI only to check their work, not replace their thinking. After over 20 years of experiential learning across Asia, our team at The Global Institute for Tomorrow understands productive struggle. Working on real problems in unfamiliar contexts builds judgment, resilience and problem-solving skills that are immune to chatbot competition. Companies may in the end decide to bond junior employees with a contract under which professional development is provided in return for a promise to remain for a minimum period of time. The education system needs urgent reform. Long-form reading of printed books, a practice in danger of being abandoned, strengthens empathy, systems thinking and imagination in ways screen-based consumption cannot. Universities must rethink assessments. Testing for reasoning rather than polished output would reward the cognitive capabilities the labour market demands. It is beyond the scope of young professionals to reshape the job market. But like every generation, they must adapt, even if this time is more structural. The choice is between automation’s comfort and deliberate friction, which builds capability. Those who consciously choose “productive struggle” will be the ones employers hire. They will need support to make that choice. But they must make it.

How recent graduates entering the job market can outperform AI

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}As automation erodes cognitive and interpersonal skills, those who embrace the struggle of enhancing offline reasoning will thrive

As automation erodes cognitive and interpersonal skills, those who embrace the struggle of enhancing offline reasoning will thrive