In researching how artificial intelligence is changing work in Southeast Asia, I kept returning to one contradiction: people can believe that AI will remove jobs and, at the same time, believe that it will create new ones.
That combination of concern and optimism appears across Singapore, Malaysia and Indonesia. Workers may recognise the potential benefits of AI while remaining uncertain about whether their own roles, skills and opportunities will remain relevant.
Much of the public discussion focuses on the final outcome: how many jobs will disappear? But the human impact can begin much earlier. When employees do not know which tasks will change, whether their role will continue or whether they will have a fair opportunity to learn new skills, uncertainty itself can become a source of strain.
The stress can begin before a job is lost
Researchers use the term job insecurity to describe a person’s sense that their employment, or valued aspects of their job, may be under threat. It is not the same as receiving a redundancy notice. However, the absence of a confirmed job loss does not make the concern psychologically unimportant.
A systematic review and meta-analysis of longitudinal studies concluded that perceived employment insecurity is likely to have an adverse effect on mental health. Earlier meta-analytic research has also associated job insecurity with anxiety, depression, emotional exhaustion and lower life satisfaction.
Recent research has started to examine this relationship specifically in the context of AI. A 2025 longitudinal study followed 403 employees in South Korean organisations. It found no direct relationship between organisational AI adoption and employee depression. Instead, it identified an indirect relationship through job insecurity: AI adoption was associated with poorer outcomes when it increased employees’ concerns about the security of their jobs. The relationship was weaker in organisations with stronger corporate social responsibility practices.
One study cannot explain every workplace or national context. Nevertheless, the finding points to an important distinction. AI is not automatically harmful to worker wellbeing. Much depends on how technological change is introduced, explained and managed.
Fear and optimism can coexist
The regional picture is more complicated than a simple choice between being optimistic or pessimistic about AI.
In the Ipsos Predictions 2026 survey, 76% of respondents in both Indonesia and Singapore, and 72% in Malaysia, thought it was likely that AI would lead to many jobs being lost in their country.
At the same time, 68% of respondents in Indonesia, 67% in Malaysia and 60% in Singapore thought AI would lead to many new jobs being created.
These beliefs are not necessarily contradictory. A person may expect some jobs to disappear and others to emerge while remaining unsure about where they will fit within that transition.
The survey also requires careful interpretation. It asked whether respondents thought AI would cause job losses in their country—not whether they personally expected to lose their own job. Ipsos also notes that the online samples in Indonesia, Malaysia and Singapore were more urban, educated and/or affluent than their general populations and should be understood as reflecting the views of their more connected population segments.
The results should therefore not be presented as evidence that three-quarters of workers expect to be personally displaced. They do, however, show that AI-related job disruption is highly visible in the public imagination.
Malaysia: “Affected” does not mean “eliminated”
Malaysia illustrates why the language used to describe workforce change matters.
Malaysia recorded 42,807 reported job losses between January and 12 June 2026. According to data cited by Human Resources Minister Datuk Seri R. Ramanan, 40.85% of these losses were associated with business closures and company downsizing.
These figures should not be treated as evidence that AI caused all—or even most—of the reported job losses.
TalentCorp’s Impact Study of AI, Digital and Green Economy on the Malaysian Workforce examines 949 job roles across ten sectors. Its published findings estimate that approximately 620,000 employees within those sectors could be significantly affected by technological and economic shifts. The study also identifies 60 emerging roles and maps transition pathways for 84 highly affected roles.
Being “affected” is not the same as being made redundant. Affected roles may involve different tasks, require additional skills or develop into new forms of work. Presenting every affected role as a future job loss can unnecessarily intensify anxiety and obscure the opportunities that may also emerge.
The need for preparation is nevertheless real. PwC reported that 57% of Malaysian employees expected their jobs to be significantly affected by technology and AI within three years, while only 63% said they had access to the learning and development resources they needed.
For employers, the lesson is not simply to tell workers to upskill. They must explain what is expected to change, which skills will be needed and how employees will be supported in developing them.
Indonesia: Optimism depends on access
Indonesia shows the strongest coexistence of concern and optimism among the three countries in the Ipsos findings. While 76% of respondents expected AI to lead to job losses, 68% also expected it to create new jobs.
Indonesia’s comparatively high use of AI may provide part of the context.
PwC’s Global Workforce Hopes and Fears Survey 2025 included 812 respondents from Indonesia. Sixty-nine per cent said they had used AI in their work during the previous year, while 16% reported using generative AI daily.
Daily users were more likely than infrequent users to report benefits. Ninety-six per cent of daily users reported productivity improvements, compared with 75% of infrequent users. Eighty-two per cent of daily users reported greater job security, compared with 63% of infrequent users. These are self-reported associations and do not prove that using AI more frequently automatically causes greater productivity or job security.
Access to learning opportunities was also uneven. Sixty-four per cent of non-managers said they had the resources they needed for learning and development, compared with 89% of senior executives.
This gap matters. Employees with the least influence over organisational decisions may also have less access to the resources they need to adapt to those decisions.
Optimism about AI may therefore depend on whether workers have opportunities to use it, understand it and benefit from it. When those opportunities are distributed unequally, the same technology may feel like a useful tool to some employees and a threat to others.
Singapore: Change is entering an already pressured workplace
In Singapore, concerns about AI are appearing within a workforce that is already reporting substantial pressure.
ManpowerGroup’s Global Talent Barometer 2026 surveyed 515 workers in Singapore between September and October 2025. Thirty-nine per cent anticipated the possibility of losing their job within six months, while 58% feared that AI-driven automation could replace them within two years.
The same survey found that 53% experienced significant daily stress and 72% had recently experienced burnout. These figures do not establish that AI caused the reported stress or burnout. They show that concerns about automation are entering a workplace environment in which many employees already feel strained.
The findings also reveal a gap between confidence in the present and uncertainty about the future. Eighty-five per cent of respondents were confident that they had the skills required for their current role, and 52% regularly used AI at work. However, confidence in using the latest technology stood at 69%, and more than half of the surveyed workers had not recently received training or mentorship.
Employees may therefore feel capable of doing their jobs today while remaining unsure about what their organisation will expect from them tomorrow.
This distinction is important. Asking employees to experiment with AI while maintaining the same workloads and performance expectations can make learning feel like an additional demand rather than a source of support. Training needs to be accompanied by time, guidance and realistic expectations.
What employers can do
Say what is changing—and what is not
Organisations should distinguish between the automation of a task, the redesign of a role and the elimination of a job. These outcomes are not interchangeable.
Employees should be told what decisions have already been made, what remains uncertain and when further information is likely to become available. Vague statements about “AI transformation” can leave people to imagine the worst.
Leaders may not have every answer, but they can still be honest about what they know and do not know.
Make learning part of the job
Upskilling is less credible when it consists only of optional online courses that employees are expected to complete outside working hours.
Learning should be connected to actual roles and future opportunities. Employees need protected time to practise, examples of responsible AI use and guidance on how new skills may lead to different tasks, projects or career pathways.
Access should also extend beyond senior leaders and employees who already work in technology. Otherwise, development programmes may widen the inequalities they are intended to address.
Give employees a role in implementation
Employees should have opportunities to ask questions, raise concerns and contribute to decisions about how AI will affect their work.
The people performing a role often understand its practical demands, exceptions and risks better than those selecting a new system. Their participation can improve implementation while giving them a greater sense of control over the transition.
Technological change is likely to feel less threatening when it is being developed with employees rather than simply imposed on them.
Track wellbeing alongside productivity
Organisations commonly assess AI through time saved, output produced or costs reduced. These measures do not reveal whether employees are experiencing heavier workloads, reduced autonomy, role confusion or anxiety about their future.
AI initiatives should therefore be evaluated using both operational and human measures. Relevant questions include whether employees understand how their role is changing, feel confident using the technology, have access to learning and trust the organisation to communicate honestly.
Managers should also know how to respond when employees raise concerns and where to direct those who require additional workplace or professional support.
A transition people can see themselves in
The evidence from Singapore, Malaysia and Indonesia does not support a simple story in which workers either welcome AI or reject it.
People can expect AI to displace some jobs while believing that it will create others. They can use AI and recognise its benefits while remaining anxious about whether their skills will be valued in the future.
My main takeaway from researching these countries is that the most important question is not whether AI is simply “good” or “bad”. It is whether workers can see a credible place for themselves in the future their organisations are building.
A fair technological transition requires more than new tools and training programmes. It requires clear communication, time to learn, meaningful employee participation and attention to wellbeing.
When those conditions are present, AI is more likely to feel like a change people can navigate—not a threat they can only wait for.