We are counting courses completed, certifications issued, and AI tools deployed. What we are not counting — not carefully enough, anyway — is who is being left out of all of it.
There is a version of the AI upskilling story that sounds almost aspirational. Governments are investing. LinkedIn feeds are full of people announcing their latest certifications. Organisations are proud of their AI pilots. And yet, when I look across the teams I work with — in healthcare, in financial services, in growth-stage companies across Singapore and the region — I see a pattern that does not make it into those announcements: the people most at risk of being displaced are the least likely to be in the room where the reskilling is happening.
That is not a skills problem. It is a structural one.
The numbers are not reassuring
Across Southeast Asia, nearly 46 percent of firms have begun scaling AI — but among SMEs, which employ the majority of the region's workers, adoption stands at roughly 15 percent, even in Singapore. Upskilling investment tends to follow adoption investment. If smaller businesses are not yet deploying AI seriously, they are also not building the capability to work alongside it.
The workers most exposed to displacement are concentrated in service, sales, and clerical roles. Women are nearly twice as likely to face AI-driven disruption as men, who tend to occupy roles in manufacturing and manual labour that are less immediately affected. In South Asia, women are up to 40 percent less likely to own a smartphone — which means the access gap precedes the skills gap.
The skills needed for jobs across Southeast Asia are expected to change by 72 percent between 2016 and 2030 — nearly double the rate of the prior decade. In agriculture alone, up to 5.7 million jobs could vanish by 2028. Administrative roles, disproportionately held by women, carry high automation risk. The pace of change is accelerating precisely in the sectors least served by current upskilling investment.
What the ground actually looks like
I work across two contexts simultaneously — a regional corporate environment and an independent consulting practice — and the pattern I observe in both is consistent.
When I joined a regional healthcare organisation late last year, one of my earliest observations was how differently AI was landing across functions. Strategy, marketing, and data teams were actively experimenting — prompting, iterating, building workflows. Administrative and operational staff — many of whom had spent years developing deep process knowledge — had received a single briefing and a link to a policy document.
This is not unique to healthcare. Across the consulting engagements I run, the same pattern repeats: AI literacy investment tends to follow seniority and function, not exposure to risk. The people with the most agency over their own learning — and the most time and tools to pursue it — are pulling further ahead. Everyone else is waiting for a programme that has not been designed for them yet.
The pipeline is not helping either. Only three percent of employers believe higher education is adequately preparing graduates for an AI-driven future, according to Singapore's Digital Education Council. If the entry point is broken, the divide compounds from day one.
What I find harder to quantify — but equally real — is the erosion of mid-career confidence. When a professional who has spent a decade building expertise in research, analysis, or client communication watches AI replicate those outputs in seconds, the psychological cost does not show up in a workforce report. But it shapes whether they lean into reskilling or quietly disengage.
Three shifts that would actually help
- Design access around exposure, not enthusiasm. The people who most need AI capability building are rarely the ones proactively seeking it. Organisations need to map their highest-risk roles and bring the learning to those people — not wait for those people to find their way to the learning.
- Make it applied, not aspirational. Giving someone a subscription to an online course library is not upskilling. The organisations I have seen make real progress embed learning into actual work — short, contextual, tied to a real task or output. A customer service team learning to use AI to triage and respond more effectively builds lasting capability. A team completing a generic AI module does not.
- Count who is missing. ASEAN's high levels of workforce informality mean that equitable access to digital infrastructure cannot be assumed. Within organisations, the equivalent question is simpler: when we run capability programmes, who attends? Who does not? Why?
The divide is a choice
Around 164 million workers across Southeast Asia could be affected by AI, with women and younger workers disproportionately impacted. The upskilling conversation cannot remain a story about the already-ready getting more ready.
The quiet divide in AI upskilling is not inevitable. But it will not close on its own. It will close when organisations stop measuring participation and start measuring impact — and when the people designing these programmes ask, from the beginning, who is not yet in the room and why.
