Malaysia Says 10% of Jobs Face Change as AI Tests Skills and Employment Plans

One in 10 jobs faces automation or a change in function

About 10% of existing jobs in Malaysia face automation or changes to their functions as artificial intelligence, digitalisation and the green economy reshape employment, according to Economy Minister Akmal Nasrullah Mohd Nasir. The warning comes alongside a target to raise skilled employment from 30.1% of the workforce in 2025 to 35% by 2030.

In a written Dewan Rakyat reply to Jimmy Puah Wee Tse, the member for Tebrau, Akmal cited a November 2024 Talent Corporation Malaysia study. He said the remaining 90% of jobs could stay relevant through skills adaptation. That distinction matters: the ministry is describing exposure to automation and restructuring, not predicting that one in 10 workers will lose their jobs.

The government sees the shift as an opportunity to move towards industries that generate greater economic value and more skilled employment. Its challenge is to make training and new job creation move fast enough for workers whose existing duties become less valuable or disappear.

Several figures help explain the scale of that challenge, but they measure different things. TalentCorp identifies 620,000 employees likely to be affected within 10 sectors. A separate academic analysis estimates 320,000 jobs at risk under a generative AI scenario, rising to approximately 680,000 with deeper adoption of systems that can execute entire workflows.

Those numbers cannot be treated as interchangeable estimates of national job losses. They differ in sector coverage, technology assumptions and whether they count changing tasks, obsolete roles or possible displacement. Reading them together shows why Malaysia needs both a strategy for creating new employment and practical routes for people already working.

What TalentCorp actually measured

TalentCorp's workforce impact study examined 949 job roles across 10 sectors responsible for approximately 60% of Malaysia's gross domestic product. It projects that about 18% of employees within those sectors, approximately 620,000 people, will be significantly affected by technological and economic changes.

The sectors include aerospace, chemicals, electrical and electronics, energy and power, food manufacturing and services, global business services, information and communications technology, medical devices, pharmaceutical manufacturing, and wholesale and retail trade. Their economic contribution was RM933 billion in 2023.

The 18% figure applies to the workforce in those sectors, while Akmal's 10% statement concerns existing jobs more broadly. The official study page does not provide a calculation reconciling the two percentages. Neither percentage, by itself, establishes how many people will become unemployed.

There is also a difference in how the 620,000 figure has been described. In a June 2025 institutional article, Steven Sim described those jobs as facing a high risk of replacement or obsolescence. TalentCorp's study page uses the broader description of employees expected to be significantly affected. Restructuring a role and eliminating it are different outcomes.

The study offers more than a count of exposed workers. It maps career transitions for 84 highly affected roles and identifies 864 basic and occupation specific skills. Those details could help employers connect a disappearing task to a realistic next role, rather than send workers into training without a destination.

Why services and clerical work face particular pressure

A 2 June 2026 ISEAS Perspective paper by financial economics practitioner Lim Kok-Tiong places Malaysia's services sector at the centre of AI exposure. It employs approximately 10 million people, or 65% of total employment, and contributes roughly 60% of GDP.

Applying an Asia Pacific generative AI exposure estimate of 22% to that workforce, the paper calculates that up to 2.2 million service jobs could experience some degree of AI assistance or task change. This is a broad extrapolation, not a count of jobs already automated.

Generative AI produces material such as text, images and software code. In workplaces, it can draft correspondence, summarise documents and process information. A job becomes exposed when some of its tasks overlap with those capabilities; exposure does not mean the system can reliably perform the whole job.

Research by the Institute of Strategic and International Studies Malaysia and the World Bank estimates that 92% of clerical support roles are highly exposed. It also describes task overlap affecting approximately 4.2 million Malaysian workers. That larger figure should not be read as 4.2 million clerical jobs destined for removal.

World Bank economist Alyssa Farha Jasmin, a coauthor of the research, explained that the findings identify tasks that could be automated depending on employer decisions, rather than inevitable layoffs. The same software can help an employee handle more work or allow a company to reduce staffing. How it is deployed determines which outcome follows.

Women and the middle of the labour market

The clerical findings point to an uneven burden: women account for 84% of clerical staff. Changes to administrative work could therefore affect women disproportionately, even if the national employment total looks relatively stable.

Age patterns are more complicated than a simple warning that the youngest workers face the greatest exposure. The ISIS Malaysia and World Bank research finds workers aged 15 to 24 most represented in the lowest exposure category, while those aged 25 to 64 are most represented in the highest exposure category.

Lower exposure is not necessarily an advantage. Younger workers may be concentrated in manual jobs requiring few qualifications, while automation reduces the administrative and cognitive positions that could otherwise offer a route into better paid work.

Khazanah Research Institute's technology and labour analysis, published on 18 September 2024 by Dr Mohd Amirul Rafiq Abu Rahim and Goh Hao Yi, identifies routine work, particularly in the middle skills category, as especially exposed. Malaysian citizens hold 90% of jobs in that category.

The institute cites estimates that 11% to 54% of Malaysian jobs could be significantly affected by automation over two decades. That wide range concerns automation more broadly and a much longer period than the immediate policy discussion. It also notes that nearly half of graduates are overqualified for their first job, citing the Higher Education Ministry's Graduate Tracer Study. More qualifications alone will not solve a shortage of suitable employment.

When AI assistance becomes job replacement

Lim's ISEAS analysis separates two possible stages of adoption. In the first, generative AI assists workers and automates highly exposed tasks. In the second, agentic AI extends automation into more complex workflows. Agentic systems can carry out sequences of actions, rather than simply answer an individual prompt.

The paper estimates approximately 320,000 jobs at risk in a scenario limited to generative AI and approximately 680,000 if agentic adoption deepens. Against a services workforce of 10 million, those figures represent about 3.2% and 6.8%, respectively. They are scenarios based on assumed adoption intensity, not forecasts.

Its stress calculations suggest unemployment could rise from a baseline of 3.5% to approximately 5.5%, or towards 6.3% under a further assumption that labour force participation falls from 70% to 60%. These are conditional modelling results, not official unemployment projections.

The paper also cites an October 2024 Human Resources Ministry statement that approximately 300,000 jobs had been lost to automation and AI since 2020. It notes that no annual breakdown was given, making it difficult to separate recent AI effects from earlier automation and disruption during the pandemic.

Akmal separately cited a Stanford University study dated September 2026, titled How Does AI Change Labor Demand? Evidence from 41 Countries. According to his parliamentary reply, it found that generative AI was shifting demand towards experienced workers, with automation of basic tasks threatening opportunities for new entrants. That concern connects job displacement with another risk: fewer junior roles through which future skilled workers acquire experience.

Training money is substantial, but outcomes matter

Malaysia has committed several streams of funding to skills development. TalentCorp's November 2024 announcement said Human Resources Minister Steven Sim Chee Keong had announced support through a RM3 billion ministry fund, using levies, credits, scholarships and matching grants.

In his June 2025 article for the World Economic Forum, Sim, writing about his work as human resources minister, put annual spending on education and training related to skills at RM10 billion. Approximately 30% came from a statutory levy on private corporations, equivalent to about RM3 billion.

TalentCorp has also been reported to have allocated RM1.6 billion for digital and AI skills training, with courses free for Malaysian workers. These sums describe different funding arrangements and periods. They should not be added together as though they were separate parts of one new AI budget.

Akmal said 17 MyMahir Future Skills Talent Councils, each chaired by an industry representative, had been established. MyMahir provides information on job trends, required skills, career routes and training opportunities, allowing course design to follow actual employer needs.

Sim said workers trained under council skills programmes received a mean wage 12% above the national average in the first year. That is a useful comparison, but it does not show a 12% pay increase caused by training. Establishing that would require information about participants' wages before training and comparable workers who did not participate.

What the latest participation figures show

Deputy Human Resources Minister Khairul Firdaus Akbar Khan told the Dewan Rakyat that 74,332 people had participated in Department of Manpower upskilling and reskilling programmes as of 31 August. HRD Corp had approved 19,084 training places with RM47.67 million in financial assistance as of August 2026.

The programmes target workers in the lower and middle skills categories. Their subjects include AI, the Internet of Things, automation, drones, green technology, electric vehicles, hydrogen technology and entrepreneurship. The Internet of Things connects devices and sensors so that equipment can exchange data and be monitored.

Khairul Firdaus also announced that the maximum training fee funding under SOCSO's Skim LINDUNG Kerjaya would rise from RM3,000 to RM7,000 effective 1 October 2026. That is an increase of RM4,000, more than doubling the previous ceiling.

TalentCorp's MyMahir AI Tour and JAMx programmes had benefited more than 7,971 participants. These participation and approval figures cannot simply be combined into a total of unique workers, because overlap is not specified. Nor do they establish how many participants obtained new jobs.

Khairul Firdaus cited Farm Fresh Berhad's estimated RM14 million investment in AI monitoring of product expiry dates, which reduced customer complaints by 30%. The example shows a measurable operational benefit, but no accompanying employment figure establishes whether staffing increased, decreased or remained unchanged.

Can employers turn training into useful work?

Employer preparation remains uneven. An Alibaba Cloud survey found that 68% of respondents believed companies were lagging in cloud computing and AI adoption. The PwC Malaysia Corporate Directors Survey 2024 found that 23% of Malaysian CEOs planned to integrate AI into workforce and skills strategy.

Other survey findings point to limited training preparation. EY's 2023 Work Reimagined Survey found that 22% of Malaysian companies were preparing to train workers for AI. The PwC survey found that 12% prioritised generative AI training and 4% expressed concern about related job losses. The surveys cover different populations and periods, so they are indicators of preparedness rather than a single comparable trend.

Farid Basir, chief people officer at MBSB Bank and a Malaysian Employers Federation council member, advocated keeping people central to automation decisions. His core recommendation was:

AI should augment, not replace people.

Farid argued that people remain essential for empathy, interpreting complex situations, innovation and values. The ISIS Malaysia and World Bank research recommends stronger unemployment insurance, earlier AI education and ethics, skills such as critical thinking and complex judgment, and incentives for employers to retrain staff.

Lim's proposed AI social contract similarly calls for internal redeployment, transition routes, adaptable safety nets and greater transparency when AI substitutes for human tasks. These are proposals rather than announced commitments. Their common point is that a training course is only one part of a successful employment transition.

New roles, university plans and the 2030 target

TalentCorp identifies approximately 60 emerging roles expected to grow in importance over three to five years, including AI engineers, sustainability specialists and engineers working with biological production processes. Sim said 70% of those roles were in AI and digital technologies, equivalent to about 42 types of role, not 42 individual jobs.

Malaysia also targets 500,000 employment opportunities through planning related to AI. Akmal has pointed to technical and vocational education and training as a way to help workers adapt, with the Human Resources Ministry responsible for delivery and the Economy Ministry setting policy direction.

The Higher Education Ministry aims to produce 30,000 graduates with AI skills between 2025 and 2030 through 30 specialised programmes at 11 public universities. Minister Zambry Abdul Kadir said 455 academic programmes had been reviewed as of 30 September and that specialised AI courses follow the Malaysian Qualifications Agency's Programme Standards: Artificial Intelligence.

Dedicated faculties include AI at Universiti Teknologi Malaysia, AI and cybersecurity at Universiti Teknikal Malaysia Melaka, and intelligent computing at Universiti Malaysia Perlis. The ministry also plans to include AI elements in 80% of bachelor degree courses at public universities.

Under the 13th Malaysia Plan, increasing skilled employment from 30.1% to 35% requires a rise of 4.9 percentage points. Jobs in the middle skills category accounted for 57% of employment in 2025, while those in the lowest category accounted for 12.9%. The target therefore involves changing the composition of employment, not merely producing more AI specialists.

The sequence of commitments and what is unresolved

Workforce policy spans the 13th Malaysia Plan, the New Industrial Master Plan 2030, the National Energy Transition Roadmap, the National Semiconductor Strategy and the Malaysia Digital Action Plan 2030. Akmal said the digital plan, launched by Prime Minister Anwar Ibrahim on 29 June, supports an adaptable and inclusive workforce as Malaysia pursues its AI nation ambition.

The main milestones show how the response has developed:

• 18 September 2024: Khazanah Research Institute published its analysis of technology, skills mismatch and automation exposure.
• 18 November 2024: TalentCorp launched its first study covering 10 sectors and announced support through a RM3 billion ministry fund.
• 10 June 2025: Sim outlined annual skills spending, emerging roles and plans for 65,000 National Training Week courses for ASEAN citizens that June.
• 2 June 2026: ISEAS published Lim's scenarios for generative and agentic AI adoption.
• August 2026: HRD Corp training approvals reached 19,084 places with RM47.67 million in assistance.
• 1 October 2026: The announced SOCSO training fee funding ceiling rises to RM7,000.
• 2030: Malaysia targets a 35% skilled employment share and 30,000 graduates with AI skills across the 2025 to 2030 period.

TalentCorp initially planned to extend its research to 12 additional sectors in 2025. Its study page now describes Phase 2 as in progress, without a firm publication date. Results from that expansion could help clarify exposure beyond the original sectors.

Other gaps remain consequential: there is no annual breakdown for the reported historical losses, no demonstrated reconciliation of the national and sector percentages, and no evidence that new opportunities will match displaced workers by location, pay or qualifications. The practical test is whether workers can move into suitable employment before changes to their existing roles leave them without an income.

Key Points

• The Economy Ministry says about 10% of jobs face automation or changes in function, not necessarily elimination.
• TalentCorp projects significant effects for 620,000 employees, or 18% of workers within 10 studied sectors.
• ISEAS estimates 320,000 to 680,000 jobs at risk under different AI adoption scenarios, not forecasts.
• Clerical work is highly exposed, with women accounting for 84% of clerical staff.
• Malaysia aims to raise skilled employment from 30.1% in 2025 to 35% by 2030.
• Training participation is expanding, but the figures do not yet establish how many workers secure new employment.
• TalentCorp's next research phase remains in progress without a firm release date.