Enterprises are moving beyond experimentation to full-scale adoption of AI agents in the coming year. Deloitte’s “The State of AI in the Enterprise” report indicates that the number of companies with more than 40% of projects in production is set to double in six months. Closer to home, the same report highlights that 32% of respondents in Singapore have moved 40% or more of their AI pilots into production, higher than the global average of 25%.
After rounds of pilots and prototypes, organisations will expect AI agents to start driving tangible business outcomes. Where scaling, governance, and cost control have been obstacles, connecting AI agents to real-time, governed data, and integrating them across business workflows could support full-scale AI adoption. Organisations are moving from using AI as a passive consultant to deploying it as an autonomous agent capable of executing “high-stakes” workflows across industries and business functions.
Anthropic’s latest capabilities, deploying agent teams capable of handling complex, multi-step tasks independently, are a step towards operationalising AI agents at scale. AI agents are now capable of helping financial services firms detect and assess risk faster, automate compliance reporting, and deliver more personalised customer interactions. In telecommunications, this means modernising network operations, streamlining customer lifecycle management, and enhancing service delivery.
Capability, however, does not guarantee outcomes. The organisations that will extract the most value from these advances are not those with access to the most powerful models, but those with the strongest data foundations underlying them.
Break down data silos before they calcify
Fragmentation across the data estate hinders any organisation from maintaining consistency, governance, and control. Organisations risk having different departments choose their own tools, run their own proof of concepts (POCs), and deploy solutions independently. Much like the early days of business intelligence, we are beginning to see AI silos forming within enterprises.
At the same time, Deloitte’s report reveals that agentic AI usage is poised to rise sharply in the next two years, yet only one in five companies has a mature model for governance of autonomous AI agents. A global study by Cloudera, “The Evolution of AI: The State of Enterprise AI and Data Architecture,” reports that only 2% of organisations in Singapore have access to all of their organisation’s data for AI initiatives.
Without a unified view, data visualisation can be incomplete or misleading, often resulting in ineffective decision-making. Enterprises must therefore prioritise data architectures that break down silos, enforce consistent governance, and provide a single source of truth for analytics and AI.
Maintaining control through private AI
Human oversight is still essential to ensuring data quality and governance, providing a strong foundation for enterprises to flexibly deploy multiple AI tools and models to optimise workflows. To achieve this regardless of where the data resides and without vendor lock-in, organisations are increasingly exploring “private AI” architectures. Leveraging platforms that are secure by design helps enforce data residency and access controls.
Deploying models on-premises also allows organisations to retain control over their data and AI models, supporting compliance and security throughout the AI lifecycle.
Embed governance as a foundation, not an afterthought
As AI agents take on greater autonomy, the risks of overlooking governance increase. Organisations must abide by data sovereignty requirements, ensuring that data remains within the appropriate jurisdiction and aiding compliance with local and international regulations. Data exposure to external entities should be limited to reduce the risk of breaches. Traceability is key to ensuring that AI models remain accountable.
With AI agents becoming increasingly embedded in regulated industries, explainability is increasingly treated as a compliance requirement, giving organisations visibility into how decisions are made, what data was used, and where outputs can be audited.
As organisations become spoilt for choice with the emergence of new models and agents, the importance of integrating AI into their broader data environment becomes clearer. Underpinned by strong data foundations, standardised metrics, and sustainable governance, enterprises that keep pace with innovation in this manner will be better positioned to capture value from AI adoption.
