Agentic AI is being hyped as the fourth big wave of artificial intelligence, but its implications for business are still up for debate. Capgemini Research Institute estimates agentic AI may release up to US$450 billion of economic value by 2028. However, adoption is still in its infancy: only 2% of businesses have scaled its application, and faith in AI agents is already beginning to erode.
That tension – high potential but low deployment – is what Capgemini’s new research explores. Based on an April 2025 survey of 1,500 executives at large organisations in 14 countries, including Singapore, the report highlights trust and oversight as important factors in realising value. Nearly three-quarters of executives said the benefits of human involvement in AI workflows outweigh the costs. Nine out of ten described oversight as either positive or at least cost-neutral.
The message is straightforward: AI agents perform optimally when alongside humans, not on autonomous pilot.
Early moves, gradual progress
Around a quarter have deployed agentic AI pilots, and only 14% have begun implementation. Deployment remains in the planning phases for most. The report terms this as a growing gap between intent and readiness, now one of the key impediments to economic value capture.
The tech isn’t theoretical – real-world uses are beginning to turn up, and an example is a personal shopping assistant that can look up items based on a request, create product descriptions, respond to questions, and add items to a cart via voice or text input. These resources usually fall short of actually completing financial transactions for security purposes, but they already simulate many of the actions of a human assistant.
This leaves larger questions about the function of traditional websites. If AI will take care of such tasks as searching, comparing, and making purchases, will humans need to find websites directly themselves? For those who get overwhelmed or confused on busy websites, an AI-managed interface could provide a more streamlined, easier alternative.
Defining agentic AI
In an effort to drill through the hype, AI News interviewed Jason Hardy, Hitachi Vantara’s chief technology officer for artificial intelligence, regarding how businesses in Asia-Pacific should approach the technology.
Agentic AI is software that can make choices, take action, and adapt its plan independently,” Hardy explained. “Consider it a group of domain specialists that can learn through experience, manage tasks, and act in real time. Generative AI generates content and is typically reactive to input. Agentic AI might employ GenAI within it, but its purpose is to seek goals and act in changing environments.
The differentiation – between creating outputs and pushing outcomes – encapsulates the essence of agentic AI for enterprise IT.
Why adoption is gaining momentum
Adoption is being propelled by scale and complexity, as Hardy suggests. “Businesses are overwhelmed by complexity, risk, and scale. Agentic AI is gaining traction because it does not simply analyse. It optimises storage and capacity in real time, automates governance and compliance, pre-empts failures before they happen, and reacts to security threats in real time. That movement from ‘insight’ to ‘autonomous action’ is why adoption is speeding up,” he said.
This is supported by research from Capgemini. The report discovered that though there is uneven confidence in agentic AI, early implementations are being found valuable when the technology assumes mundane but critical IT functions.
Where value is emerging
Hardy cited IT operations as the strongest use case up to this point. “Automated data classification, proactive storage optimization, and reporting for compliance save teams hours a day, while predictive maintenance and real-time responses to cyber threats cut downtime and risk,” he stated.
The reach is broader than efficiency. The capabilities enable systems to identify issues ahead of time, distribute resources more optimally, and quarantine security incidents faster. “Early adopters are already applying agentic AI to remediate incidents ahead of time prior to escalation, building reliability and performance in hybrid environments,” Hardy said.
For the present, IT is the most pragmatic place to begin: its implementation has quantifiable outcomes and is at the core of how businesses approach cost and risk management, demonstrating the implications of agentic AI in business.
Southeast Asia’s outset
Southeast Asian organisations, Hardy continued, must start with good data. “Agentic AI only delivers value when enterprise data is classified, secured, and governed correctly,” he stated.
Infrastructure also comes into play, in that agentic AI needs systems capable of supporting multi-agent orchestration, persistent memory and dynamic resource allocation. Without this building block, adoption will be narrow in scope.
IT operations can often be where many companies start, with agentic AI able to pre-empt outages and optimize performance prior to deployment to broader business functions.
Reworking core workflows
Hardy predicts that agentic AI will transform workflows in IT, supply chain management, and customer service. “In IT operations, agentic AI can forecast capacity requirements, redistribute workloads, and redistribute resources in real time. It can also automate predictive maintenance, stopping hardware failures before they happen,” he said.
Cybersecurity is also a field of promise. “With cybersecurity, agentic AI can scan for anomalies, quarantine infected systems, and invoke immutable backups in seconds, shortening response times and minimizing the harm that can be done,” Hardy said.
The capabilities go beyond proof-of-concept tests. Initial deployments already demonstrate how agentic AI can enhance reliability and resilience within hybrid environments.
Skills and leadership
Adoption will also demand new human capabilities. “Agentic AI will change the human role from execution to oversight and orchestration,” Hardy explained. Leaders must establish boundaries and track autonomous systems so they remain in ethical and organisational boundaries.
For executives, it implies reduced attention to bureaucratic matters and more to innovation, mentoring, and strategy. HR departments will have to develop governance competencies such as audit readiness and establish new frameworks for embedding agentic AI in a meaningful way.
The impact on the workforce will be unequal. The World Economic Forum says that AI may generate 11 million jobs in Southeast Asia by 2030 and replace nine million. Women and Gen Z are likely to be hit the hardest, with over 70% of women and as many as 76% of younger employees in at-risk jobs.
This underscores the imperative for reskilling, and significant investments have already been made, with Microsoft investing $1.7 billion in Indonesia and launching training programmes in Malaysia and the broader region. Hardy underscored that capacity building has to be inclusive, speeded up, and strategic
What comes next
Three years from now, Hardy expects most leaders to downplay the speed of change. “The first wave of advantages is already apparent in IT operations: agentic AI is automating jobs such as data classification, storage optimisation, predictive maintenance, and cybersecurity response, allowing teams to concentrate on higher-level strategic work,” he stated.
But the biggest surprise could be at the business and economic model level. IDC estimates AI and generative AI may contribute approximately US$120 billion to the GDP of the ASEAN-6 by 2027. Hardy envisions the implications as more extensive and quicker than many anticipate. “The suggests the impact will be much faster and more material than many leaders currently anticipate,” he stated.
In Indonesia alone, over 57% of occupation jobs are likely to be displaced or supplemented by AI, a reminder that change is not only going to happen within IT. It will slice across the way companies are organized, the way they handle risk, and the way they create value.
Balancing autonomy with oversight
The Capgemini findings and Hardy’s insights converge on the same theme: agentic AI holds huge promise, but its meaning in practice depends on balancing autonomy with trust and human oversight.
The technology may help enterprises lower costs, improve reliability, and unlock new revenue streams. But without a focus on governance, reskilling, and infrastructure readiness, adoption risks stalling.
For Southeast Asia, the issue is not whether agentic AI will gain traction, but how soon – and whether organizations can walk the line between autonomy and accountability as machines start shouldering greater business decision-making responsibility.