How Singapore Can Protect Its Middle Tier of Jobs as AI Reshapes Work

A banking career shows the pressure on the middle

At 48, banking professional Rei Teo found that experience was no guarantee of a secure next step. Her job was reshaped during a restructuring in 2024, and she left. With hiring cooling and technology changing the roles available, she struggled to find another permanent position and eventually accepted a contract role.

Her experience illustrates a concern identified by Singapore's Institute for Adult Learning (IAL): the country has too few jobs in the middle tier to absorb displaced professionals. Experienced workers can find themselves moving into temporary assignments or positions with lower pay because there are fewer openings at their former level.

IAL calls this a push down effect. Companies automate routine tasks and streamline teams using artificial intelligence, leaving fewer stable roles in the middle. Professionals seeking work compete for a smaller pool of suitable positions, while slower junior hiring puts pressure on the career ladder from the other end.

Singapore is a lead partner in an international research programme examining AI adoption in 10 global digital hubs, from the Republic to Silicon Valley. Findings so far point to a particular risk for economies with a large share of professional jobs and a thin middle tier: the jobs needed to absorb displaced workers and train future leaders may become scarcer.

As of September 20, 2025, the challenge is to protect both routes into that middle: progression for younger workers and a way back into stable employment for experienced professionals. The choices involve employers, training institutions and government, because acquiring skills is only part of the answer. Workers also need jobs in which to use them.

Why the middle matters beyond individual careers

Supervisors, team leaders, business analysts, financial controllers, product managers, clinical coordinators and senior technicians perform different tasks, but share an important function. They turn an organisation's plans into work that teams can deliver.

These workers also carry knowledge that is difficult to put into a manual: how a client's needs have changed, where a process regularly fails, or which part of a system requires special care. This is often called tacit knowledge, meaning understanding built through experience rather than written instructions. They coach junior colleagues and maintain continuity when teams change.

A thinner middle leaves senior leaders managing more detail and gives new employees fewer opportunities to learn through practice. The ladder from education to professional competence and then leadership becomes harder to climb. For Singapore, that threatens the social compact linking effort with reward and can weaken trust in employers and institutions.

Singapore has long sought to avoid a barbell economy, with many highly paid jobs at the top and many lower paid service jobs at the bottom, but too little in between. AI can help build skilled, well paid careers if it supports employees and creates specialist paths. It can also narrow those paths when companies remove the roles in which people first gain experience.

How AI changes tasks before it changes jobs

Since late 2022, banks, consultancies and technology companies have been testing AI copilots and chatbots for customer service, code generation, compliance checks, marketing copy and searches of internal knowledge. A copilot is an AI assistant that helps a person complete work, such as preparing a draft or finding relevant information.

AI adoption often starts with parts of a job. A tool might summarise documents, reconcile data or prepare a report while a person checks the result. Some tasks disappear; others grow or require new oversight. Together, these changes can alter the size and shape of teams across professional work once considered relatively sheltered from automation.

Financial services make the shift especially visible. Automated checks and document reviews are spreading through risk and compliance work. Operations teams use copilots for reporting and data reconciliation, while teams serving clients test assistants for research and service. Roles are also emerging in model monitoring, data governance and AI risk management.

Model monitoring means checking how an AI system performs over time. Data governance sets responsibilities for the information it uses. AI risk management deals with problems that can arise from its use. These activities create work requiring human judgment even as other tasks become automated.

Banking, consulting, technology and corporate services face two connected pressures. Removing junior tasks reduces opportunities to develop future professionals. Removing tasks from established roles can leave experienced employees competing for fewer jobs in the middle. Both processes weaken the same part of the workforce.

The first rung is becoming harder to reach

In early 2025, one large bank said AI could reduce its contract and temporary workforce by up to 4,000 roles over the following few years. Years earlier, the Monetary Authority of Singapore had projected that automation could transform or eliminate one in three finance jobs in the Republic.

Across marketing, software and operations, tools can draft copy, design graphics, debug code and summarise documents in seconds. Those tasks have often given junior employees their first chance to understand a profession. Removing them changes both recruitment needs and the way newcomers learn.

A Stanford University analysis of millions of US payroll records found a 13 percent relative decline in employment among workers early in their careers in the occupations most exposed to AI following the wave of generative AI adoption in late 2022. Employment among older workers in the same occupations was stable or increased. The adjustment occurred more through reduced hiring than lower pay.

Generative AI produces material such as text, images or computer code. The employment pattern identified in the US points to pressure on the first rung of the career ladder, where employers can reduce recruitment as tools take over work previously assigned to beginners.

Singapore surveys reflect anxiety among younger workers. Roughly half of Gen Z respondents expect AI to cause major disruption to their roles, and six in 10 say they are already seeking extra training. Only about a third of companies provide structured AI learning. Some young professionals are moving towards skilled trades, seeking work they consider less vulnerable to automation, with greater stability and clearer progression.

A training trap affects young and experienced workers

Junior employees need practice to develop judgment. When AI supplies the first draft, their work may look polished while their understanding remains shallow. Senior colleagues can spend more time revising it, while beginners have fewer chances to learn why an answer is right or wrong.

Training becomes weaker if an entry role centres on prompting a tool without mastering the underlying craft. The immediate output may arrive faster, but the pipeline of people ready for independent responsibility becomes thinner. That eventually affects the supply of capable professionals in the middle.

Experienced workers face a different problem. Knowledge and professional networks can protect some older employees, consistent with the stability found among experienced workers in the Stanford analysis. Yet professionals in their 40s and 50s who leave during restructuring can struggle to return as permanent positions tighten and contract assignments expand.

Teo's experience shows why a stronger middle requires two forms of support at once: a steady inflow of trained juniors and structured routes back into suitable work for people changing careers or recovering from displacement.

The developments span several years:

• Since late 2022, generative AI adoption has accelerated, changing tasks across professional services and providing the starting point for the Stanford employment analysis.
• In 2024, Teo left after her banking job was reshaped during restructuring and subsequently struggled to secure permanent work.
• In early 2025, a large bank said AI could trim up to 4,000 contract and temporary roles over the next few years.
• By September 20, 2025, Singapore faced the linked tasks of supporting displaced professionals and preserving opportunities for younger workers to gain experience.

Managers determine how AI enters the workplace

Company leaders decide whether AI adoption centres on reducing headcount or on helping employees do more valuable work. Managers in the middle translate those decisions into everyday practice. They need time, resources and backing from senior leaders to redesign workflows, retrain colleagues and build trust in new tools.

Basic AI literacy is the starting point. Employees need to understand data, the limits of models and privacy responsibilities. Experiments with limited risk, followed by shared lessons and early successes, can help teams identify where the tools are useful.

Customer service, finance, human resources and operations offer opportunities to pair AI assistance with human oversight. Teams drawn from different functions can help successful pilots expand and resolve bottlenecks in existing processes. Training needs include model monitoring, prompt engineering, workflow design and managing organisational change.

Incentives matter too. Managers rewarded mainly for reducing headcount have a different reason to act from managers rewarded for developing junior colleagues and redesigning jobs. Protecting the middle requires treating people development as part of successful AI adoption.

Policy can rebuild routes into skilled work

Singapore used traineeships during the pandemic to help graduates get started. A proposed permanent national apprenticeship route for digital and professional work would extend that approach, with paid placements lasting 12 to 24 months across finance, healthcare, public service, manufacturing and media.

Such placements would need structured mentorship, skills checklists and regular feedback. AI tools would form part of the work, while trainees also learn the judgment needed to check and apply their output. Countries with dual track systems combine education with workplace learning to support the move into employment and keep youth unemployment low. Scale and continuity matter to making this approach work.

Workers changing careers need funding and time as well. Career Conversion Programmes, grants for training at work and SkillsFuture credits can be directed towards roles in data, operations and service as AI changes demand. Expanded subsidies for people aged 40 and above could make it easier for employers to recruit experienced candidates from related sectors.

Short qualifications that build towards a diploma or degree provide another route. These stackable credentials let workers demonstrate progress in stages and give employers a shared way to recognise capabilities.

Public support for AI adoption could also carry commitments to employees. Grants for redesigning jobs could require trainee places, records of process changes and learning resources shared within the company. Wage support could help employers redeploy staff into teams needing human judgment, including AI risk, customer success and product operations.

A workforce trained in AI safety and governance, supported by clear standards, could meet new demand for oversight and offset some automation risk. These proposals focus on connecting technology investment to jobs and progression, rather than expecting training alone to create openings.

The public service provides a practical example

All public officers will be required to complete an AI literacy course, while public service leaders are being asked to use AI thoughtfully in daily work. One in three officers already uses an internal assistant, and thousands of custom bots have been built to automate routine tasks.

Deputy Prime Minister Gan Kim Yong has called on leaders to set the tone by embracing AI responsibly and encouraging teams to use tools where they improve service and governance. The public service approach combines a common foundation of learning with applications developed for everyday tasks.

For large institutions, that connects leadership decisions with practical experience. Employees need both an understanding of the technology and opportunities to apply it with appropriate responsibility for the results.

What employers can change inside their organisations

Employers can widen the middle by retaining experience, making progression visible and giving displaced staff a chance to move internally. The practical measures extend beyond buying software or offering an introductory course.

• Use AI first to remove repetitive work, then reinvest the time in training, client care and product quality.
• Create structured apprenticeships for knowledge work, pairing every junior employee or trainee with a mentor, a skills curriculum and tracked progress towards greater responsibility.
• Record practical knowledge in playbooks, checklists and internal reference pages so that experience survives changes in staffing.
• Link managers' bonuses to developing juniors who can work independently and to successful internal promotions.
• Advertise internal roles openly and recruit according to skills so displaced employees can move between functions.
• Create specialist careers in data, operations, design and risk, with clear pay bands and status equal to management paths.
• Assign responsibility for model risk, data quality, privacy and bias, and train employees for those duties.
• Share early successes and failures so employees understand where AI works and where it falls short.

These measures address different parts of the same problem. Mentorship maintains the supply of future professionals, internal mobility preserves experience, and specialist careers create progression without requiring every capable worker to become a manager.

Workers need evidence of skills and judgment

Workers can strengthen their position by combining knowledge of a field with AI fluency and human skills. The aim is to identify a problem, choose suitable tools and deliver a useful result.

• Learn how copilots work, where they fail and how to prompt them effectively. Practise checking and debugging their output.
• Develop knowledge of a sector's data, regulations, workflows and customer problems, giving AI skills a practical foundation.
• Learn to map processes, write clear handovers and present findings. Explaining and persuading remain useful across roles.
• Build data judgment by reading dashboards, designing simple experiments and tracking results connected to revenue, risk, cost or customer outcomes.
• Keep a portfolio showing improvements in speed, quality or service achieved with AI. Concrete results make capabilities easier to assess.
• Consider stackable credentials in data analytics, cloud tools, cybersecurity basics or AI safety, alongside professional experience.
• Join professional communities and volunteer for projects across functions where AI is being tested, creating opportunities to learn through participation.

Professionals changing careers can look for related roles that value their existing judgment. Operations managers can move towards process improvement using AI. Relationship managers can combine customer success work with analytics. Compliance officers can move into oversight of model risk.

Describing experience through problems solved and results delivered helps make those connections visible. Demonstrating how AI extends that experience gives an employer something more specific than a general claim of familiarity with new tools.

How Singapore can measure progress

Singapore can assess the health of its middle tier through employment stability, access to training and actual movement into more responsible work. Course participation is useful, but job searches, promotions and successful transitions show whether learning connects with opportunity.

Measures proposed for tracking that progress include:

• The ratio of permanent to contract roles in professional and technical occupations.
• The median time displaced professionals aged 40 to 55 spend searching for work.
• The share of entry job advertisements offering structured training or apprenticeships.
• Rates of internal promotion from junior positions into the middle tier across major sectors.
• The number of career conversions completed each year through national programmes.
• Participation and completion rates for AI literacy courses in the public and private sectors.
• The creation of governance roles, including model risk and data stewardship, that offer new opportunities in the middle.

Tracking these measures would help employers, workers and policymakers see whether Singapore is rebuilding the career ladder: giving beginners real practice, helping experienced people return to stable work and creating skilled roles around the technology.

Key Points

• IAL identifies a push down effect in which displaced professionals face too few stable jobs at their previous level.
• AI can reduce junior learning opportunities while changing established roles, putting pressure on both routes into the middle tier.
• A Stanford analysis found a 13 percent relative employment decline among workers early in their careers in the US occupations most exposed to AI, while older workers were stable or gained employment.
• Roughly half of Singapore's Gen Z survey respondents expect major AI disruption, and six in 10 are seeking training. About a third of companies provide structured AI learning.
• Paid apprenticeships, support for career conversion and employment commitments attached to AI grants are proposed ways to strengthen progression.
• All public officers will be required to complete AI literacy training. One in three already uses an internal AI assistant.
• Employers can protect experience through mentorship, internal moves and specialist careers, while workers can demonstrate a combination of domain knowledge, AI skills and judgment.