Agent AI has been the talk of the next major wave of artificial intelligence, but its meaning for businesses has not been resolved. The Capgemini Research Institute estimates that Agent AI can unlock as much as US$450 billion by 2028. However, recruitment is still limited. Only 2% of organizations are expanding their use, and trust in AI agents is already beginning to slip.
That tension – potential but less unfolding – is what Capgemini’s new research is exploring. Based on an April 2025 survey of 1,500 executives from large organizations in 14 countries, including Singapore, the report highlights trust and oversight as key factors in achieving value. Almost three-quarters of executives said the benefits of human involvement in AI workflows outweigh the costs. Nine out of ten have been described as false surveillance as positive or at least neutral.
The message is clear. AI agents work best when paired with someone who is not left in the autopilot.
Early steps, slow progress
About a quarter released Agent AI pilots, but only 14% of people moved to implementation. For the majority, deployment is still in the planning stage. This report describes this as a growing gap between intent and preparation. It is currently one of the main barriers to achieving economic value.
This technology is not theoretical, real-world applications are beginning to emerge, with one example being a personal shopping assistant that can search for items based on a particular request, generate product descriptions, answer questions, and place them in your cart using voice or text commands. These tools are usually not lacking in completing financial transactions for security reasons, but they already replicate many of the capabilities of human assistants.
This raises greater questions about the role of traditional websites. If AI can handle tasks like search, comparison, or purchase, do you need to navigate directly through your online store? For those who find busy websites overwhelming or difficult to navigate, AI-driven interfaces may offer simpler and more accessible options.
Definition of Agent AI
To get through the hype, AI News We spoke with Jason Hardy, Chief Technology Officer of Artificial Intelligence at Hitachi Vantara about how companies in the Asia-Pacific region should think about technology.
“Agentic AI is software that lets you decide, act and refine your strategy on its own,” says Hardy. “Think as a team of domain experts who can learn from experience, coordinate tasks, and operate in real time. Generated AI creates content and responds to prompts. Agent AI may use genai in it, but the job is to pursue a purpose and take action in a dynamic environment.”
The distinction between output generation and operational outcome captures the meaning of Agent AI in Enterprise IT.
Why adoption accelerates
According to Hardy, recruitment is driven by scale and complexity. “Companies own at complexity, risk and scale. Agent AI is catching up to do more than analytics. This is why it adopts a shift from “insight” to “automatic operational behavior” that optimizes storage and capacity, automates governance and compliance, responds to security threats in real time.”
Capgemini’s research supports this. This study found that trust in agent AI is heterogeneous, but early deployment has proven useful when technology takes on routine but essential IT tasks.
Value is emerging
Hardy pointed to IT tactics as the most powerful use case ever. “Automized data classification, proactive storage optimization and compliance reports save teams time every day, while predictive maintenance and real-time cybersecurity response reduce downtime and risk,” he said.
The impact exceeds efficiency. The feature means that the system can detect problems before escalates, allocate resources more effectively, and include security incidents more quickly. “Early early users are already using Agent AI to actively repair incidents before escalating, enhancing reliability and performance in hybrid environments,” Hardy added.
For now, it remains the most practical starting point. The deployment provides measurable results, is central to how companies manage both costs and risk, and demonstrates the implications of agent AI operations.
Southeast Asia starting point
For Southeast Asian organizations, Hardy said the number one priority is to get the data right. “Agent AI only provides value when enterprise data is properly classified, protected and governed,” he explained.
Infrastructure is also important. This means that Agent AI needs a system that can support multi-agent orchestration, persistent memory, and dynamic resource allocation. Without this foundation, recruitment is limited.
Many companies may choose to start with IT operations where Agent AI can preempt outages and optimize performance before deploying it to a wider business function.
Reshaping the core workflow
Hardy expects Agent AI to rebuild workflows, supply chain management and customer service. “In IT operations, Agent AI can also redistribute resources in capacity needs, rebalancing workloads, real-time, and can also automate predictive maintenance and prevent hardware failures before they occur,” he said.
Cybersecurity is another realm of commitment. “In cybersecurity, Agent AI can detect anomalies, isolate affected systems, cause immutable backups in seconds, reduce response times and reduce potential damage,” Hardy noted.
Functions are not limited to proof of concept tests. Early deployment already shows how agent AI can enhance reliability and resilience in hybrid environments.
Skills and leadership
Recruiting also requires new human skills. “Agent AI shifts human roles from performing to monitoring and orchestration,” Hardy said. Leaders need to set boundaries, monitor autonomous systems, and ensure that they remain ethical and organizational limits.
For managers, this change means focusing on management tasks and focusing on mentoring, innovation and strategy. HR teams need to build governance skills such as preparatory audits and create new structures to effectively integrate agent AI.
The impact of the labor force is uneven. The World Economic Forum predicts that AI will be able to create 11 million jobs and drive away 9 million people by 2030. Women and Generation Z are expected to face the most sudden confusion in more than 70% of women and up to 76% of young workers in roles vulnerable to AI.
This highlights the urgency of reskills, highlighting that major investments are already underway as Microsoft has committed $1.7 billion in Indonesia and rolls out training programs in Malaysia and the wider region. Hardy emphasized that capacity building should be comprehensive, prompt and strategic.
What’s coming next
Looking three years from now, Hardy believes that many leaders underestimate the pace of change. “The first wave of profits is already seen in IT operations. Agent AI is freeing up to automate tasks such as data classification, storage optimization, predictive maintenance, cybersecurity response, and other tasks, allowing teams to focus on high-level strategic work,” he said.
But the bigger surprise may be at the economic and business model level. IDC Project AI and Generation AI could add approximately US$120 billion to ASEAN-6’s GDP by 2027. “This impact suggests that it will become much faster and more material than many leaders expect right now,” he said.
In Indonesia, it is expected that more than 57% of employment roles will be enhanced or destroyed by AI, and this reminds us that conversion is not limited to that. It reduces the structure of your business, how you manage risk, and how you create value.
Balancing autonomy and surveillance
Capgemini’s findings and Hardy’s insights converge on the same theme. Agent AI holds a big promise, but the actual meaning depends on balancing autonomy, trust and human surveillance.
The technology could help businesses reduce costs, improve reliability and unlock new revenue streams. However, recruitment risk stalls without focusing on governance, reskilling and infrastructure preparation.
For Southeast Asia, the question is not whether agent AI will take hold, but how quickly and whether companies can balance autonomy and accountability as they begin to take more responsibility for business decisions.
(Photo: Igor Omilaev)
See: Beyond Acceleration: Rising Agent AI
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London. The comprehensive event is part of TechEx and will be held in collaboration with other major technology events. Click here for more information.
AI News is equipped with TechForge Media. Check out upcoming Enterprise Technology events and webinars here.
