APAC enterprises are funding AI agents, but control and accountability remain the harder test

Enterprise AI spending in APAC is moving into a more demanding phase, as companies shift attention from experiments to systems that can act across workflows, data, infrastructure, and security environments.

A new special report from Omdia, Informa’s technology research and advisory group, found that 42% of surveyed organisations are allocating US$1 million or more to AI agents over the next 12 months. The report, Staying Ahead in the AI Era, was released ahead of Asia Tech x Enterprise, which returns from 20 to 22 May 2026 at Singapore EXPO.

AI agents are moving beyond assisted workflows

The level of planned spending suggests agentic AI is gaining enterprise traction faster than Generative AI did at comparable stages of market development. Unlike tools that mainly support content generation or decision-making, AI agents can initiate actions, coordinate workflows, and execute tasks without continuous human involvement.

That creates a different adoption challenge. The report frames agentic AI as a break from the long-standing business software model where humans decide and technology assists. For enterprises, the issue is less about access to AI capability and more about how work, oversight, and accountability are redesigned around systems that can act.

Human roles are also expected to shift towards exception handling, judgement, and oversight. This will require companies to define who remains responsible when automated systems make recommendations, trigger actions, or move work between teams. For industries with regulated processes or complex approval chains, those operating rules may become as important as the AI models themselves.

ROI becomes the condition for bigger spending

Omdia’s findings also show that enterprises are still applying financial discipline to AI investment. More than three-quarters of surveyed businesses, or 82%, said they are willing to commit larger budgets if measurable and defensible business value can be delivered.

“The decisions enterprise leaders make in the next 18 months will define their competitive position,” said Joyce Wang, Event Director for Asia Tech x Singapore at Informa. “We created this special report with Omdia to serve as a strategic orientation tool for the region. By mapping out these critical shifts, from agentic AI to quantum readiness and physical AI, we are ensuring that when delegates arrive at the Singapore EXPO this May, they are ready to transform these innovations into measurable business impact. At Asia Tech x Enterprise, our goal is to equip leaders with the comprehensive insights needed to not just track the market, but to confidently lead it.”

The investment case is becoming harder to separate from deployment readiness. AI agents require integration with enterprise applications, reliable data access, governance processes, and security controls. These costs can rise quickly if use cases are not tied to clear operational gains.

The report also connects the move towards production AI with infrastructure change. It describes traditional IT infrastructure across Asia as moving towards “AI Factories”, heavy industrial facilities designed to continuously produce intelligence. This places AI adoption closer to an infrastructure planning issue, especially as enterprises demand higher compute capacity and more reliable deployment environments.

Sovereign AI faces practical limits

Sovereign AI remains a priority across the region, but Omdia’s report presents it as difficult to achieve under a strict definition. A recent Omdia survey found that 64% of enterprises support sovereign AI to protect local data and align with national priorities.

The challenge is the depth of capability required. According to the report, strict AI sovereignty is highly demanding and remains unachievable for most countries outside the United States and China. This leaves APAC markets with a practical trade-off: they want greater control over data and AI capability, but few can build the full stack alone.

The report indicates that APAC nations will increasingly need strategic regional alliances and big tech capabilities to support national AI ambitions. For governments and enterprises, sovereignty becomes a question of how much control can be retained across infrastructure, data, models, talent, and policy.

Security risks widen as AI deployment deepens

As AI systems gain access to more enterprise workflows, security becomes a larger part of the deployment agenda. The report says organisations need to balance four areas: Security with AI, Security by AI, Security against AI, and Security for AI.

This reflects the dual role of AI in cybersecurity. AI can strengthen detection and response, but it can also increase the speed and scale of threats. Protecting AI systems themselves is also becoming part of enterprise security planning, rather than a separate technical issue.

Quantum risk adds another layer to that planning. Omdia cites a 2030 deadline when quantum computing is projected to break current encryption methods such as RSA and AES. Almost one third of organisations, or 32%, are already actively exploring quantum-resistant protections.

The report also points to physical AI as part of the next enterprise technology cycle. Humanoid robotics is expected to mature in 2026, with South Korea investing in its robotics ecosystem and Taiwan emerging as a key contract manufacturer for humanoid robotics vendors.