Editor
Dinesh Raj chevron_right
Malaysia has spent much of this year bracing for an AI shock to its workforce. In May, TalentCorp warned that up to 697,000 jobs could be significantly affected by artificial intelligence, digitalisation and the green economy within three to five years if workers do not reskill. The figure has anchored headlines and boardroom slides ever since. A new set of data from Google suggests the picture is more complicated, and in some ways less alarming, than the fear implies.
Google's first ATLAS report, published in late July, draws on roughly 15 million anonymised interactions with its Gemini assistant. It found that AI has already reached about 70 percent of occupations, covering some 90 percent of United States employment. Yet the depth of that use is shallow. The average worker turns to AI for only about a fifth of their tasks, and fewer than one in ten interactions involve handing a job over to the machine outright. Most of the time, people use AI as a collaborator: to research, draft, troubleshoot, iterate and learn.
That distinction, between reaching a job and replacing it, matters for how Malaysia reads its own numbers. The ATLAS data also showed AI use spreading well beyond desk-bound professionals, with technicians and tradespeople turning to conversational tools for real-time diagnostics and on-the-fly learning. The through-line is augmentation rather than automation: the tools are changing how work gets done long before they remove the worker doing it.
Malaysia's own forecasts, read carefully, point the same way. Alongside the 697,000 roles it flagged as high risk, TalentCorp estimated that around 2.2 million jobs would be augmented by AI over time, and identified 120 emerging roles, most of them in AI itself and in the green economy. The risk, in other words, is concentrated among workers who do not adapt, not spread evenly across the economy.
Local employers are already living in that in-between state. Microsoft's 2026 Work Trend Index found Malaysian workers broadly ready to use AI, but readiness is not the same as relief. Studies of AI at work, here and abroad, have found that time saved on one task tends to be reinvested into new ones, with a majority of employees reporting heavier rather than lighter workloads. The machine is not clocking off; it is raising the bar for what one person is expected to produce.
There are reasons to treat Google's findings with caution. The data comes from Google's own product and reflects self-directed use rather than deliberate workforce restructuring, and it captures an early moment in a fast-moving technology. Adoption that looks like harmless assistance today can harden into genuine job redesign tomorrow, especially in routine cognitive work.
For Malaysia, the practical lesson is neither panic nor complacency. If the near-term reality is augmentation, then the return on reskilling is unusually high right now: the same tools that threaten a worker who ignores them make a worker who learns them measurably more productive. That is the window the 697,000 figure is really describing, and it is one the country still has time to use.
Images courtesy of Vitaly Gariev and Arlington Research on Unsplash.
