The 26% Problem: Why ASEAN’s AI Ambitions Keep Hitting the Same Wall

Ong Khai Wei’s digital footprint leaves a trail of high-level execution. His Medium blog runs technical tutorials on building AI agents with Python and debugging CrewAI deployments. In early 2026, he took on the role of workshop lead and judge for “Hack for Cities,” a competition designed to spark innovation among SME students. IBM supports the program by providing the technical mentorship and judging required to turn student ideas into viable prototypes. He is the kind of executive who still writes code on weekends, and that matters because everything he says about why ASEAN enterprises struggle to unlock value from data comes from watching what actually breaks at the engineering layer, not from a slide deck.

Ong is IBM’s field chief technology officer and technology community leader based in Singapore. He arrived from Malaysia 17 years ago and has spent most of that time in the regional enterprise technology ecosystem, working on conversations that span boardrooms in Jakarta to data centers in Kuala Lumpur. He has heard many reassuring things over the years.

Most of them don’t survive contact with the data.

The number that changes everything

According to IBM’s 2025 Chief Data Officer Study, 81% of chief data officers worldwide now say their data strategy is integrated with their technology roadmap. This is a sharp jump from 52% just two years ago. The C-suite has finally stopped treating data like a byproduct and started treating it like a strategic asset.

Here’s the number that complicates that progress: only 26% of those same CDOs believe their data capabilities can actually support new AI-driven revenue streams.

You’ve briefed the board and bought the platform. You’ve then integrated the roadmap. And still, nearly three in four of your peers cannot make their data do what the AI strategy promises. In a winner-takes-most AI market, this isn’t a gap. It’s a liability with a countdown.

The platform fallacy

Ong has watched the same assumption take hold in boardrooms from Jakarta to Kuala Lumpur. It usually sounds like this: “We have comprehensive data tools and platforms that centralize all the data for AI.”

It’s an understandable place to stop. It’s also not the right one.

“Having the technology is essential to make data ready for AI adoption,” Ong says, “but it does not mean it is AI-ready to deliver the expected insights and value demanded by the business.”

Strip away the dashboards and the architecture diagrams, and most ASEAN enterprises are still wrestling with three persistent problems. Data silos: multiple source systems, incompatible schemas, no shared taxonomy. Data quality: missing fields, duplicates and records that haven’t been reconciled in years. Data governance: the persistent inability to answer the basic question of who is allowed to see what—and to enforce that answer at scale.

IBM’s study identifies the same five structural barriers globally: accessibility, completeness, integrity, accuracy and consistency. They are not frontier problems, but the same data challenges enterprises have been documenting since the data warehouse era. What has changed is not the problem. It is the price of leaving it unsolved.

Join the executive roundtable “Building AI-Ready Data Foundations: Turning Data Strategy into Scalable Outcomes” on May 13, 2026, in Singapore to hear Ong and other IBM experts and regional data practitioners talk about current challenges. Register here.

When agents meet unready data

Some 80% of CDOs are now building datasets to train AI agents. Yet 79% admit they are still working out how to govern and scale them.

Sit with that gap for a moment.

Ong has seen the consequences stack up in real deployments across the region. Agents that can’t reach the data they’re designed to consume because integrations were never fully completed. Agents that surface information to users who were never supposed to see it, because access controls built for human users don’t apply to autonomous software executing thousands of decisions per minute. Agents that return confident, wrong answers because the underlying data was incomplete, which customers catch.

“The adoption of enterprise-grade Agentic AI is more complex than CDOs expected,” Ong says.

The requirements list is long and largely underestimated: guardrails for autonomous decision-making, AI agent identity management using protocols like SPIFFE, MCP-based data integration, vector databases for unstructured content and governance frameworks that account for agents operating without human approval at every step. Most enterprises have a few of these in place. Almost none have all of them. And agentic AI deployed on top of an unready data foundation only amplifies every gap in the data it touches.

Sovereignty as a competitive architecture

In ASEAN, this data readiness challenge runs directly into something uniquely regional: 10 sovereign nations, 10 data residency regimes, 10 regulatory philosophies and 10 definitions of what it means to keep data safe. A Singapore enterprise operating across Indonesia, Thailand and Vietnam isn’t navigating one legal environment. It’s architecting around constraints that shift with every election cycle and every new bilateral technology agreement.

European data leaders negotiate with a single regulator in Brussels. ASEAN data leaders negotiate with the continent.

IBM’s study found that 82% of CDOs now treat data sovereignty as a critical pillar of their risk management strategy. The most forward-leaning among them have made a subtler move: they treat regulatory compliance as competitive architecture.

Ong’s example is financial services. Banks across Southeast Asia have operated under heavy regulatory constraints since before the internet existed. And the best of them have used that constraint as an accelerant, all while maintaining the trust that customers extend to institutions that handle their data with demonstrable care. “Consumers have confidence in financial institutions because they know their data is safe and protected while accessing it at their fingertips,” Ong says.

That trust premium is not incidental to the product. It is the product. Every sector in ASEAN is watching the playbook. The question is whether they act on it.

Singapore went first

In January 2026, Singapore unveiled what is now the world’s first comprehensive governance framework specifically designed for agentic AI, announced at the World Economic Forum and published by the Infocomm Media Development Authority. The framework addresses how organizations should assess risk before deploying agents, where human accountability must be preserved and how to bound the autonomous actions an AI agent can take within enterprise systems.

Ong is precise about what this is and isn't. “Those are not regulations or mandates. They are guidelines,” he says. But guidelines from Singapore carry weight across the region. The city-state has spent years positioning itself as ASEAN's governance laboratory, and CDOs from Manila to Kuala Lumpur follow what IMDA publishes.

The smarter organizations are using the framework as scaffolding. They are mapping the IMDA principles directly onto internal data and AI policies, strengthening their governance posture and building a brand that is auditable and trustworthy across borders. Regulatory clarity, taken seriously, becomes a head start.

Treat data like a product

Ong’s core prescription has nothing to do with budget or tooling. It has to do with how organizations fundamentally conceive of data itself.

Stop treating data as records sitting in repositories. Start treating it as a product.

Data-as-a-Product means assigning clear ownership, defining who the users are, publishing service-level agreements and measuring quality the way engineering teams measure uptime—continuously, against a published standard. IBM’s report frames this as the shift from “data readiness” to “decision readiness.” Ong distills it to a phrase: data as a first-class citizen.

The reframe is not cosmetic. Once data is treated as a product, governance becomes a feature rather than a tax. Metadata management, lineage tracking, quality metrics and audit trails stop being compliance checkboxes and start being things you build because your users demand them. AI agents stop guessing at data provenance and start consuming versioned, trusted, contractually defined data assets.

With 78% of CDOs already naming proprietary data as their top strategic differentiator, the business case is clear. As general AI capabilities commoditize, the only sustainable advantage will be data that your competitors cannot replicate. That means data that is governed, trusted and decision-ready.

The room worth being in

On May 13, IBM is convening a closed executive roundtable in Singapore, “Building AI-Ready Data Foundations: Turning Data Strategy into Scalable Outcomes.” A select group of Asia’s senior data leaders will work through the exact questions this piece raises. Every participant walks out with a personalized Data Strategy Blueprint.

Go. Not because the frameworks aren’t already public—they are. But because the playbook for closing the 26% readiness gap hasn’t been written yet. It takes shape in rooms where peers are willing to talk honestly about which governance challenges remain unsolved, what broke in production and what they actually did about it.

That conversation doesn’t happen in webinars or keynote halls. It happens when a small group of people with real accountability for real data problems sit down and exchange hard-won intelligence.

The gap is closing. Slowly, for most. Faster, for a few. Which side of that divide your organization lands on will depend on decisions made in the next 12 months, not the next five years.

Image credit: iStockphoto/SARINYAPINNGAM