Agentic AI is being hailed as the next transformative wave of artificial intelligence. Capgemini Research Institute projects it could unlock as much as US$450 billion in value by 2028. Yet adoption remains modest—just 2 percent of organisations have scaled its use, and many executives remain cautious.
Capgemini’s survey of 1,500 global executives shows that oversight and trust are central to unlocking value. Nearly three-quarters say the benefits of keeping humans in the loop outweigh the costs, while nine out of ten view oversight as either neutral or positive for efficiency. The signal is clear: agentic AI delivers best when guided by human judgment.
Early pilots, lingering hesitation
Around a quarter of organisations have tested agentic AI through pilot projects, but only 14 percent have gone further into real implementation. For most, deployment remains at the planning stage. This gap between intention and readiness is fast becoming a barrier to capturing potential economic benefits.
Practical examples are emerging. AI-powered personal shopping assistants can now search for products, generate descriptions, answer questions, and even place items into carts. These tools stop short of handling payments but already mirror many tasks performed by human assistants—raising questions about whether traditional website browsing will remain necessary.
What makes agentic AI different
Jason Hardy, CTO for Artificial Intelligence at Hitachi Vantara, describes agentic AI as “software that can decide, act, and refine its strategy on its own.” Unlike generative AI, which reacts to prompts, agentic AI actively pursues goals in dynamic environments. Hardy frames it as the difference between producing outputs and driving outcomes.
This distinction matters for enterprises: agentic AI is not just about generating content, but about making decisions and taking action in real time.
Why enterprises are paying attention
For Hardy, the real driver is scale and complexity. “Enterprises are drowning in complexity, risk, and scale. Agentic AI is catching on because it optimises storage and capacity, automates compliance, anticipates failures, and responds to security threats in real time,” he explained. The transition from insight to autonomous action is what sets agentic AI apart.
Capgemini’s findings support this perspective: early deployments are proving useful in routine but critical IT tasks, where efficiency gains are immediate and measurable.
IT operations as the proving ground
So far, IT operations have been the strongest proving ground. Automated data classification, proactive storage optimisation, compliance reporting, and predictive maintenance save hours of manual work. Real-time cybersecurity responses reduce downtime and mitigate threats. These capabilities give organisations both cost savings and resilience benefits.
Hardy notes that some enterprises are already using agentic AI to remediate incidents before they escalate, strengthening reliability in hybrid environments.
Southeast Asia’s challenges and opportunities
For Southeast Asian enterprises, adoption starts with getting the basics right. Hardy stresses that “agentic AI delivers value only when enterprise data is properly classified, secured, and governed.” Infrastructure readiness is equally critical, requiring support for multi-agent orchestration, dynamic resource allocation, and persistent memory.
Given these prerequisites, many organisations in the region will start with IT operations, where results are tangible, before expanding into areas like customer service or supply chain optimisation.
Reshaping workflows and roles
Beyond IT, agentic AI is expected to reshape workflows across industries. In supply chains, it could anticipate capacity needs and rebalance resources. In cybersecurity, it can detect anomalies, isolate systems, and initiate backups within seconds.
The shift also changes human roles. Rather than executing tasks, workers will move toward oversight, orchestration, and strategy. Managers will need to set boundaries, mentor teams, and monitor systems to ensure ethical and reliable outcomes.
Workforce disruption and reskilling
The impact on jobs in Southeast Asia will be significant. The World Economic Forum forecasts 11 million new roles created by AI in the region by 2030, alongside nine million jobs displaced. Women and Gen Z are expected to be most affected, with up to 76 percent of younger workers in roles vulnerable to automation.
Governments and companies are already responding. Microsoft, for example, has pledged $1.7 billion in Indonesia and is launching training programs across Malaysia and the wider region. Hardy emphasises that reskilling must be inclusive, rapid, and strategic to avoid leaving segments of the workforce behind.
Looking ahead
Hardy believes many leaders are underestimating the speed of change. IDC projects that AI and generative AI could add US$120 billion to ASEAN’s GDP by 2027. In Indonesia alone, more than half of job roles are expected to be disrupted or augmented.
The transformation will extend beyond IT into how businesses structure operations, manage risks, and create value. Agentic AI’s impact will be felt not just in efficiency gains, but in reshaping economic models themselves.
Balancing autonomy with trust
The research is clear: agentic AI has vast potential, but its success depends on careful governance. Enterprises in Southeast Asia—and globally—must balance autonomy with accountability.
Agentic AI could reduce costs, boost reliability, and open new revenue streams. But without trust, oversight, and strong data governance, adoption risks stalling. The region’s challenge is not whether agentic AI will take hold, but how quickly—and whether leaders can keep people at the centre as machines take on more responsibility for business decisions.
