Asia Pacific’s AI Readiness Gap Has a Hidden Cost

• The AI readiness gap in SEA and across the region is less about ambition and more about the infrastructure debt enterprises have been quietly accumulating for years.
• New IDC research finds 90% of APAC organisations have experienced failed modernisation initiatives, with poor data quality sitting at the root of both delays and outright failures.

Ask most enterprise IT leaders in Malaysia or Singapore what their biggest AI challenge is, and you’ll hear something about talent, or budget, or picking the right model. What you’re less likely to hear–but what the data keeps pointing to–is the infrastructure they inherited and never got around to fixing.

A new IDC InfoBrief commissioned by MongoDB, titled Modernising Legacy: Winning in the Age of AImakes that uncomfortable reality hard to ignore. Surveying 1,400 organisations across eight Asia Pacific markets, including Australia, China, India, Singapore, South Korea and Hong Kong, the April 2026 report finds that the AI readiness gap in this region isn’t a technology problem at its core.

It’s a data problem, one compounded by years of deferred decisions about legacy systems. The numbers are blunt. Some 95% of respondent organisations reported experiencing project delays. Nine in ten had encountered failed modernisation initiatives. In both cases, poor data quality appeared consistently as a root cause. Not budget. Not headcount. Data.

Legacy architecture is the quiet tax on every AI project

The report identifies a specific pattern worth sitting with: 43% of Asia Pacific organisations say their existing architecture makes it impossible to build new applications without extensive modernisation first. That’s not a minor technical constraint. That’s a structural veto on AI ambition, regardless of what’s been budgeted for models or compute.

IDC’s framing is pointed. Legacy relational databases, the report argues, are “too rigid, costly, and slow for today’s requirements.” Their rigidity creates what researchers describe as data debt–siloed, redundant, outdated information that quietly undermines model performance and drives up costs.

The warning that follows is not hypothetical: IDC predicts that by 2027, organisations that fail to address this data debt will face AI failure rates 50% higher than peers who do. For Malaysia and the broader SEA region, this matters. Southeast Asia has shown genuine momentum in AI adoption–the Dell and Intel-backed IDC research published earlier this week found SEA outpacing the Asia Pacific average in AI PC deployment, with employees in the region saving an average of 2.09 hours per day through AI. The endpoint and compute story is real. But compute without clean, accessible data to work with is expensive theatre.

The gap between leaders and everyone else is already widening

What makes the MongoDB-IDC report worth reading beyond its statistics is how clearly it maps the divergence already happening. IDC segments respondents into a “Leaders Cohort” and a “Mainstream Cohort”, and the distance between them is significant.

Leaders, which include not just cloud-native companies but traditional industries like manufacturing and construction, generate 71% of their revenue from digital sources today. The Mainstream Cohort sits at 23%. That’s a threefold gap, and it correlates directly with how seriously these organisations have treated modernisation as a continuous discipline rather than a one-off project.

Some 58% of Leaders have multiple programmes running concurrently to address legacy constraints. William Lee, senior research director for service provider and core infrastructure research at IDC Asia Pacific, puts it plainly: “High-quality, integrated data is the essential fuel that determines the accuracy and performance of an AI application, making modern data architecture a foundational element of any AI strategy.”

Thorsten Walther, managing director of CXO Advisory at MongoDB, adds that the research makes a clear commercial case: “Strategic modernisation unlocks AI opportunities and supports a significant increase in revenue. The leaders across the region are showing what’s possible when organisations ditch rigid, siloed legacy systems and move to AI-ready data platforms.”

The Bendigo Bank example cited in the research is instructive. The Australian institution cut the development time required to migrate a core banking application off a legacy relational database by up to 90%, at a fraction of traditional migration costs, using AI-assisted tooling.

The rework problem nobody budgets for

What this research collectively suggests is that the dominant conversation around AI costs–focused on model pricing, GPU access, and licensing–is missing the bigger line item. The real cost is the rework: going back into broken data pipelines, brittle integrations, and systems that were never designed to handle the unstructured data that AI actually runs on.

The AI readiness gap, in other words, isn’t closing itself. And for organisations across Malaysia and Southeast Asia that have invested in the visible parts of AI transformation while deferring the structural groundwork, the bill is coming due.