If you ask a chief digital officer or chief data officer for Asia Pacific to do the math on this, the numbers quickly stop adding up. You’ve spent years and a small fortune ensuring your data is stored where regulators want it. That means being localized, compliant with your location, and audited within an inch of your life. Then connect it to a large language model trained on someone else’s data, hosted under someone else’s laws, and governed by someone else’s values.
Congratulations. You have built a sovereign data lake that feeds a brain that is clearly not sovereign.
A 2026 Accenture survey of more than 750 senior leaders across the region puts the numbers on this issue. About 60% of Asia-Pacific (APAC) organizations apply sovereign controls to their data, but only about a quarter apply the same controls to the AI models that use the data to make decisions. Analysts have begun to refer to this as the “information sovereignty gap.” This is not a rounding error, this is a hole in the floor of the AI era big enough to swallow a compliance department.
Edmund Wong is sitting in the front row of the gap. He became Orange Business’ president of Asia Pacific in April 2026, stepping up after nearly six years managing the region’s operations and taking over a claim that sounds almost contrarian in an industry obsessed with model benchmarking. In other words, the hard part of enterprise AI was never the models.
“The operating environment for companies in the Asia-Pacific region is more volatile than in previous years,” Wong told CDOTrends. Rapidly accelerating technology cycles and disruptive business models are forcing companies to face what he calls a “cruel double obligation” to modernize quickly while tightly controlling security, privacy, and cost. Under that pressure, most companies do what seems prudent at the time. That means when problems arise, we add the best tools for cloud, connectivity, and cybersecurity to solve today’s fires without fixing tomorrow’s wiring.
That is the trap of “postponing” in this industry. We’ve been debating its demise for years, but it’s not dead yet. Wong’s version of the post-mortem is specific. With siled telemetry data, security blind spots, inconsistent policy enforcement, and what he calls “swivel chair operations,” teams bounce between a dozen vendor dashboards just to figure out why something broke.
The pitch: We’re a plumbing company, not an AI company.
Orange Business isn’t trying to beat Anthropic in model or Google DeepMind in computing, and Wong isn’t pretending otherwise. “Our most defensible value is not trying to compete with hyperscalers or specialized AI vendors at every layer of the technology stack,” he says. “They bring world-class platforms, models, computational power, and massive innovation.” Fine, leave it to us.
Orange flies the flag: connectivity, cloud, cybersecurity, data governance, and now AI, all run as one operating model rather than five separate contracts. Think like a structural engineer, not a technology vendor. No one hires an architect to admire steel beams. They hire it so the building doesn’t collapse when an AI floor is added on top.
This proposal flies in the face of the actual regulatory situation in the region. Rather than a single set of rules, it means dozens of countries are improvising and creating their own rules in real time. “Trusted AI starts with a trusted foundation,” Wong argues. For multinational CDOs, that means having “the right data architecture, secure hosting options, clear access controls, local data residency where necessary, and governance that allows them to operate across borders without losing sight of local obligations.” Orange frames this as “trust by design” based on EU AI law and its own internal ethics charter. It’s a very European answer to a very Asian problem, and depending on which regulator you ask, it’s either a strength or an irony.
Skeptics ask: Integrators or just well-connected resellers?
Any CDO who has attended a vendor pitch that uses the term “ecosystem” will recognize the skepticism it invites. Orange name-checks Microsoft, AWS, Google, Nvidia, Snowflake, and Databricks. A fair question is whether that makes Orange a true integrator or just a reseller with connections to other people’s platforms.
Wong is undaunted by that. “Being ecosystem-driven doesn’t mean ‘intermediating’ other companies’ platforms,” he says. “The value is not in adding another layer of complexity. The value is in making the ecosystem work for the customer.” Accountability is key, he argues. Orange is always working towards end-to-end results, not just implementation.
Regarding production AI, he points to the in-house implementation of Orange’s Live Intelligence assistant. Over 57,000 monthly active users, 2 hours and 48 minutes saved per employee per week, and 20,000 user-built agents since launch. This is proof that the model will work before it’s sold, he suggests. The next Live Intelligence Studio will move further into the realm of agents, AI that not only helps but acts. “That’s where AI starts to move from assistance to action,” he says.
where money stops
The most obvious thing Wong says has nothing to do with technology. That’s where his company’s work ends. “There are decisions the customer has to make,” he says. “Customers need to define their business priorities, acceptable risk levels, governance principles, and ethical boundaries for how AI is used within their organizations.” Orange lays the foundation. The CDO still owns the roof, risk, and leakage liability.
It may sound like a smaller claim than most vendors make, but it’s a more honest claim. Wong’s three-year measure of success has nothing to do with the number of projects. He wants customers to see Orange as part of their operating model, rather than the vendor they once hired. Not a bad definition of trust in a space where “sovereignty” is quietly the most contentious topic in enterprise technology. It’s not about who built the AI, it’s about who keeps answering the phone when it breaks.
Image credit: iStockphoto/Rafa Jodal
