How Singapore can outsmart AI age bias and redesign work for longevity: Opinion

Ageism is a “silent but deadly threat” to retirement security, wrote Claer Barrett in The Financial Times recently. In Singapore, that observation arrives at a pivotal moment.

Nearly 40 per cent of our resident workforce is already aged 50 and above. A low birth rate means fewer younger workers are entering behind them. At the same time, artificial intelligence is reshaping how organisations find, assess and deploy talent.

Yet, AI-powered recruitment systems risk scaling past hiring biases across candidate pipelines, before a human resource manager steps in.

The two forces, demographic longevity and algorithmic decision-making, do not simply create a problem to be patched.

They present a rare opportunity: to move beyond treating potential age bias as a discrete symptom, and instead rebuild the foundations of how we define talent, employment and the very nature of work for the AI age.

The same conditions that enabled algorithmic age bias to emerge elsewhere are beginning to appear in our own labour market. Singapore has repeatedly shown that it can turn structural pressures into competitive advantage. This is another such moment.

When algorithms encode outdated assumptions

Modern hiring tools, from applicant tracking systems to large language models, screen candidates using patterns learnt from historical data. They weigh graduation years, length of experience, career trajectory and even linguistic style. These proxies often correlate with age.

The result can be a quiet filter, which ranks capable older applicants lower than younger ones, before a human ever sees their CV.

Early warning signals are already visible. High-profile litigation in the US, including ongoing collective actions, has alleged that widely used algorithmic screening tools disadvantaged applicants aged 40 and over.

The same historical-data training methods are operating in our market. We have seen algorithmic bias amplification before with gender and ethnicity. What makes age distinctive here is scale.

As early Generation X and late baby boomers move through their 50s and 60s, the pool of experienced talent is larger than ever.

Yet job descriptions routinely feature terms such as “agile”, “digital-native” or “fast-paced” – which are soft proxies that may function as implicit age filters, whether intended or not.

A symptom of deeper design choices

Some average differences are commonly assumed, such as the idea that younger workers may adapt faster to certain high-velocity tools in specific roles.

When such generalisations are baked into training data, they shape how models score candidates even without explicit age criteria.

This creates challenges on two levels.

First, it penalises the many exceptions: older professionals who bring deep domain expertise, better judgment, stronger networks and proven resilience. Second, it becomes self-reinforcing: reduced opportunities create skill gaps that then appear to justify further bias.

Recent Ministry of Manpower data illustrated the exposure. In the first quarter of 2026, retrenchment incidence among residents aged 50 to 59 rose to 3.1 for every 1,000 employees, the highest of any age group.

Degree holders recorded the same rate, also the highest across education levels, as restructuring concentrated in professional and knowledge-intensive sectors.

The overlap is instructive: Many of those affected are precisely the experienced professionals moving through white-collar pipelines that increasingly rely on automated screening.

Re-entry rates have improved in recent quarters, yet longer transitions and potential retirement shortfalls remain real for many.

The data should not be read as direct proof of algorithmic ageism. It does, however, highlight where the structural risk is concentrated, and why treating it merely as a hiring-filter problem would be insufficient.

Lessons from earlier transformations

History offers useful parallels. In the early 1950s, British coal mines introduced new longwall mechanisation technology. Managers expected higher output. Instead, productivity stagnated, absenteeism rose and industrial relations deteriorated.

Researchers from the Tavistock Institute discovered why: The technology had been optimised in isolation. The social system of work – autonomous multi-skilled teams that had previously organised their own pace and division of labour – had been dismantled.

When those teams were allowed to redesign how the new technology fitted into human collaboration, performance recovered.

The insight crystallised as sociotechnical systems thinking: technology and the organisation of work must be jointly optimised, not treated as separate problems.

A more recent corporate example is IBM’s shift, beginning around 2014, towards a skills-based organisation.

Facing disruption from cloud computing and early AI, the company recognised that traditional-role definitions and linear career paths would leave large parts of its workforce stranded.

It rebuilt internal mobility, assessment and compensation around continuously updated skills rather than fixed job titles or tenure proxies.

AI itself was used to map skills and match people to opportunities. The result was not merely bias mitigation; it was a fundamental redesign of how talent is understood and developed.

Both cases share a pattern. A technological or demographic pressure first appeared as a discrete difficulty. Organisations that treated it only as a symptom to be patched underperformed.

Those that used the moment to revise underlying principles of work – how value is created, how people are evaluated and how learning continues – gained lasting advantage.

“Addressing potential algorithmic ageism is more than fairness towards today’s older workers. It is the chance to embed durable principles for how work will be organised, evaluated and rewarded in the decades ahead.”

Singapore’s opportunity to set the foundations

Singapore is already building many of the right elements. The National AI Impact Programme aims to equip 100,000 workers to become “AI bilingual”, fluent both in their domain expertise and in the practical application of AI to redesign workflows.

SkillsFuture and the new Skills and Workforce Development Agency are shifting the national conversation from one-off training towards continuous career health.

In July, retirement and re-employment ages rose to 64 and 69, respectively, signalling that longer working lives are expected and valued.

The Workplace Fairness Act, taking effect in end 2027, will strengthen protections against discrimination, including on the grounds of age.

The Infocomm Media Development Authority’s Model AI Governance Framework already provides a disciplined approach to high-impact AI use cases.

What remains is to connect these strands explicitly into a coherent redesign of talent systems for a multi-generational, AI-augmented economy. This requires five shifts:

• Move from proxy-heavy screening (based on signals such as graduation year, years of experience and “digital-native” language) to assessing skills, potential and verified capability. Use AI where it adds precision, but not where it accelerates historical bias.
• Keep the final holistic judgment human. Use AI narrowly for verifiable matching and logistics, after ensuring assessments are age-blind and proxies are neutralised.
• Regularly audit recruitment technology and test for disparate impact. Treat age as a core tested dimension, consistent with existing AI governance practice.
• Expand fair-hiring machinery to embed age-inclusive design. This extends the Tripartite Guidelines process already used for other protected characteristics, and ensures compliance with the Workplace Fairness Act.
• Design roles and career pathways for multi-generational teams. Deliberately pair the deep knowledge and judgment of experienced workers with the speed and pattern recognition of AI tools and younger colleagues. This turns blended teams into a productivity asset rather than a friction point.

In an ageing economy facing rapid technological change, short-term optimisation of screening speed can undermine long-term talent density and institutional knowledge.

The higher-productivity path is the one that treats longevity itself as an economic advantage.

Building for the generations ahead

Addressing potential algorithmic ageism is therefore more than fairness towards today’s older workers.

It is the chance to embed durable principles for how work will be organised, evaluated and rewarded in the decades ahead, principles that will serve the young professionals entering the market now as much as those already mid-career or later.

Singapore has consistently demonstrated that it can harness technology while safeguarding social cohesion.

As we supercharge workforce reskilling and navigate the opportunities and risks of AI, redesigning the foundations of talent and employment should sit at the centre of the agenda.

Getting this right will help ensure that our silver workforce remains a strategic asset, that multi-generational collaboration becomes a source of competitive strength and that longevity is converted into genuine economic advantage.

The algorithms and organisational systems we design today will shape the labour market of tomorrow.

Let us make sure they reflect the inclusive, high-productivity, continuously learning society we aspire to be, not merely for this generation of older workers, but for every generation that follows.

The writer is a senior industry expert with more than two decades in data science and AI 

Source: The Business Times © SPH Media Limited. Permission required for reproduction.

ADV: SAL Annual Lecture: The Rule of Law as a Foundation of Representative Democracy Tang See Chim, former Chua Chu Kang MP, dies at 94