According to Gartner, worldwide AI spending is forecast to reach $2.52 trillion in 2026, a 44 percent increase year-on-year. Yet nearly two-thirds of organizations have not scaled AI across the enterprise, even though 88 percent report regular AI use in at least one business function. The question is no longer whether organizations are using AI, but why these efforts are not translating into value at scale.
Most business leaders understand what successful AI scaling requires. Workflows need to decentralize, moving AI ownership closer to the business units making decisions. AI must be embedded into everyday operations rather than sitting alongside them. And the data feeding these systems must be clean, integrated, and reliable. These requirements appear in every AI strategy document, consultant framework, and industry report.
However, according to Alteryx’s latest research, which surveyed 1,400 business and IT leaders globally, including 175 in Singapore, fewer than one in four AI pilots have successfully scaled into production.
The foundations matter, but they are not enough on their own
The foundational requirements for scaling AI effectively are well recognized. Workflows need to decentralize, with responsibility for AI already shifting from central IT teams to individual lines of business, and expected to reach 33 percent of workflows by 2028. AI needs to be embedded into everyday operations rather than sitting alongside them, with 52 percent of leaders globally citing integration into core systems as their most pressing priority. And the data feeding those systems needs to be clean, integrated, and reliable, with 49 percent of Singapore leaders citing data quality as their top challenge.
The problem is that organizations can make progress on all three fronts, improving data quality, embedding AI into workflows, pushing ownership closer to the business, and still find that impact does not follow, because none of these foundations, individually or together, answer the harder question: when AI is operating across decentralized business units at scale, who ensures the controls agreed at the center are actually being applied at the point of decision?
The differentiator most organizations are missing
According to the Alteryx study, the 23 percent of organizations that have successfully scaled most of their AI pilots share a distinct profile. They report advanced data maturity, strong adherence to governance, and a consistent focus on transparency. They invest continuously in data quality and embed AI directly into existing workflows rather than run it in parallel.
What sets them apart, however, is not just the presence of these capabilities, but how consistently they are governed in practice.
The study also found that in Singapore, 60 percent of respondents cite centralized data governance as a missing capability, compared to 53 percent globally. That figure is worth sitting with, not because organizations lack governance intent, but because intent and execution are not the same thing.
Most organizations adopting AI have governance frameworks. What fewer have is governance that extends beyond the center: controls that are enforced at the point where AI is actually used, not just defined in policy. This is where the real gap lies.
A policy that defines approved data sources, human review requirements, and accountability structures is a necessary starting point. But a policy alone does not determine what happens at the moment a decision is made, or whether agreed controls are consistently applied across the business.
From policy to practice
Most scaling failures are not technical; they are operational. Governance that lives on paper cannot survive decentralized AI. So while most organizations have AI policies, what varies enormously is how consistently those policies are translated into practice.
This is not merely a theoretical challenge. FWD Insurance’s experience in Hong Kong illustrates what governance embedded into practice, rather than left in policy, can deliver. As one of Asia’s fastest-growing insurers operating across 10 markets, FWD faced a surge in reporting demands, regulatory complexity, and operational pressure as it scaled. Its finance and actuarial teams were dependent on manual workflows, with month-end reporting stretching across days and scenario testing involving countless handoffs.
The challenge was not a lack of governance intent; it was translating that intent into operations at scale. By deploying Alteryx and embedding governed, automated workflows directly into its finance and actuarial functions, FWD reduced reporting and scenario testing time by 95 percent, cut data preparation errors by 80 percent, and saved 10 days monthly on reporting. Critically, Alteryx One brought governance and collaboration to the forefront by enforcing controls not as a post-hoc check, but as a structural feature of how work gets done.
What FWD’s Director of Finance Transformation described as building a smarter, more agile finance function was in practice the work of making governance operational, with clean, integrated data, auditable workflows, and accountability embedded at the point where decisions are made, not just defined at the center.
Closing the governance gap
Singapore recently became the first government in the world to introduce a governance framework specifically for agentic AI. It is a meaningful step forward in signaling what responsible AI deployment should look like. Yet compliance remains voluntary. Without clearer enforcement standards, organizations are largely left to define and apply their own interpretations of what governance means in practice. The framework exists. How seriously any individual organization takes it is another matter entirely.
Decentralized AI workflows, tight integration into everyday operations, and high-quality governed data are the conditions for scaling. But conditions alone do not produce outcomes. What converts them into measurable business impact including lower costs, faster decisions, and improved performance, is governance that functions as enforcement, not aspiration.
Philip Madgwick the Regional Vice President, Asia at Alteryx, where he leads the company’s go-to-market teams across the region, driving strategic growth, sales and operations. He partners with enterprises to accelerate digital transformation through data analytics, enabling faster, more informed decision-making at scale.
With deep expertise in enterprise software, cloud adoption and advanced analytics, Philip focuses on advancing data literacy, scaling analytics capabilities and supporting the responsible adoption of AI across Asia-Pacific.
TNGlobal INSIDER publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.
Featured image: “Artificial Intelligence – Resembling Human Brain” by deepakiqlect is licensed under CC BY-SA 2.0