Why Traditional Hiring Won’t Work for APAC’s AI Roles

This article was originally published on e27.

I have stopped being surprised when a client tells me they have the budget, the use cases, and a talent plan that falls apart the moment someone actually tries to execute it.

This is the reality of the APAC technology talent market in 2026, and it is considerably more uncomfortable than most boardroom conversations acknowledge.

ManpowerGroup surveyed more than 39,000 employers across 41 countries earlier this year. For the first time in the history of that survey, AI skills have overtaken every other category to become the hardest to fill globally. Not engineering. Not cloud. Not cybersecurity. AI. And 71 per cent of employers across Asia Pacific report difficulty filling open roles right now. That figure was 45 per cent a decade ago. It has not plateaued. It has become, in ManpowerGroup’s own words, a structural feature of the regional labour market.

We built We+ Asia to connect organisations across Hong Kong, Singapore, Taiwan and Malaysia with the technology profiles they cannot find on their own. What we see on the ground matches those numbers, and then some.

The roles that organisations need have outpaced the talent pipelines that were supposed to produce them 

The profiles in demand today are not the same profiles that were in demand three years ago. Organisations are no longer simply looking for developers or data engineers. They need people who can work across engineering, data infrastructure and AI implementation simultaneously, who understand what a production environment actually requires as opposed to what a proof of concept allows, and who can hold a programme together across multiple vendors and stakeholders without losing the thread.

These profiles take years to develop. A six-month certification does not produce them. And with 1.6 million open AI positions globally against roughly 518,000 qualified candidates, the arithmetic is not going to improve quickly.

What compounds the problem in APAC specifically is that the senior profiles who do exist have more options than they did five years ago. Remote work opened access to US and European compensation for engineers based in Singapore and Hong Kong. The competition is no longer regional. An organisation in Kuala Lumpur is effectively competing with a tech company in Amsterdam for the same shortlist of candidates, and the Amsterdam company is probably moving faster.

The third pressure is the one organisations tend to discover too late 

The roles being created now did not exist in their current form two or three years ago. There is no reliable historical benchmark for salary, availability or hiring timeline when the job description itself is new. Organisations go into the search with assumptions built on a market that no longer exists, and they lose weeks before they realise the frame was wrong.

Put those three pressures together, and the result is not a temporary friction that resolves itself once hiring catches up. It is a gap that widens while the organisation waits for a process designed for a different era to produce a result it was never built for.

Deloitte’s 2026 enterprise AI report named the skills gap as the single biggest barrier to AI integration, ahead of data quality and ahead of regulatory constraints. That finding matches what we hear in every serious planning conversation we are part of across the region.

The organisations making real progress are thinking about this differently 

They stopped treating talent access as something that gets resolved after the roadmap is set. They build workforce planning into the transformation programme from the beginning, and they are genuinely flexible about the model, permanent hire, expert deployment, project-based engagement, whatever gets the right person into the work at the right moment.

The ones that are not doing this will still be searching when the window they had closes. That is not a projection. In several cases we are watching, it is already true.

Three questions worth asking before the next planning cycle closes.

+ First: Which roles on your roadmap require profiles that did not exist in their current form three years ago? Those are the ones that will take the longest to fill and cost the most to get wrong.
+ Second: What is your fallback if those roles are not filled on schedule? Most roadmaps do not have one, which means a single hiring miss can cascade across the entire programme.
+ Third: Are you building your workforce plan around what the market can actually supply, or around what your transformation theoretically requires? The gap between those two things is where programmes quietly lose six months before anyone says it out loud.

This is why implementation capacity has become as strategic as the roadmap itself 

Organisations that cannot find the profiles they need internally are increasingly turning to a different model. Not outsourcing in the traditional sense, but structured programme delivery with consulting partners who bring both the expertise and the people to run it. The demand for end-to-end implementation support has grown significantly across the region over the past 18 months, and it is not difficult to understand why. Having a strategy without the capacity to execute it is not a strategy. It is a slide deck with a timeline attached. 

That is the model at We+ Asia, and it is what separates programmes that deliver from programmes that drift. Not as a vendor at the edge of the project, but as an embedded partner that brings senior profiles who have done this before, in similar environments, across similar constraints.

The difference between a transformation programme that delivers and one that quietly reschedules its milestones every quarter is rarely the quality of the initial plan. It is whether the people running it have the experience to recognise what is actually happening on the ground, and the authority to act on it.

One example that illustrates the shift. A leading insurance company in Hong Kong came to us with a clear AI transformation ambition and a concrete problem: their business analysts were spending the majority of their time on manual analysis and user story creation, with no scalable path forward. They had completed pilots. They had functional specifications.

What they did not have was a delivery partner who could take AI-generated code from prototype to production safely, with confidence in quality, maintainability and live monitoring. That gap, between experimentation and production-grade deployment, is where most programmes stall.

We came in not as a recruiter, but as a delivery partner: embedding senior profiles who had solved exactly that problem before, in regulated environments, across comparable infrastructure constraints. The engagement is now in delivery. The shift from proof-of-concept to deployed AI is where implementation capacity becomes the real differentiator, not the ambition in the deck.