Agentic AI was associated with as much as $450 billion in potential economic value by 2028, yet Capgemini research found only two percent of organisations had scaled deployments. About a quarter had pilots and 14 percent had entered implementation.
The April 2025 survey covered 1,500 executives at large organisations in 14 countries, including Singapore. Nearly three-quarters said human involvement delivered more benefit than cost, while nine in ten viewed oversight as positive or at least cost-neutral.
Hitachi Vantara’s Jason Hardy distinguished agents from reactive content generation: agentic systems pursue objectives, act and refine strategies in changing environments. Early value appeared in IT operations, including data classification, capacity optimisation, compliance, maintenance and security response.
For Southeast Asia, the recommended starting point was classified, secured and governed enterprise data. Multi-agent orchestration, persistent memory and dynamic resource allocation also required infrastructure that many organisations had not yet established.
The development should be read in its original publication context. It described the evidence, forecasts and vendor claims available at that time; later outcomes should not be projected backwards into promises that had not yet been tested.
A useful assessment separates demonstrated functions from proposed scale. It compares the new approach with a realistic baseline, including manual work, waiting time, existing infrastructure and the cost of correcting errors.
Long-term value requires more than launch performance. Data, models, organisations and external rules change, so version records, periodic checks and a named owner are needed to detect drift and decide whether the original case still holds.
Distribution also matters. An average gain can conceal predictable weakness for a language, site, workload or user group. Breaking results down and inviting feedback from people closest to the process makes expansion decisions more defensible.
Before scaling, leaders should map the complete delivery chain. Data sources, interfaces, cloud services, suppliers and human review can each become the limiting factor, even when the most visible model performs well. That map also exposes concentration, security and recovery requirements that a product demonstration rarely shows.
Financial analysis should follow outcomes rather than activity. More generated content, processed alerts or automated steps is not automatically more value. Completed tasks, avoided incidents, lower correction effort and improved service quality give decision-makers a stronger basis for continuing investment.
Reversibility is equally important in a fast-moving market. Portable records, documented interfaces and clear stop criteria let an organisation change provider, narrow a deployment or return a decision to people when performance, pricing or policy changes. That flexibility is practical resilience, not a lack of commitment.
Adoption Lagged Far Behind Expectations
High projected value coexisted with very few scaled deployments. The gap suggested that demonstrations were easier than connecting agents safely to data, tools and accountable business processes.
The practical importance was that organisations needed bounded initial workflows with measurable outcomes rather than an enterprise-wide autonomy target. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Human Oversight Was Treated as Productive
Executives generally saw people in agent workflows as beneficial or cost-neutral. Humans could approve consequential actions, resolve exceptions and provide context that a model could not infer reliably.
The practical importance was that oversight had to be designed into queues, permissions and escalation paths instead of added as a vague promise after deployment. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
IT Operations Offered the Clearest Early Value
Classification, storage optimisation, compliance reporting and incident remediation had repeated events and observable results. Those characteristics made the domain more testable than open-ended strategic work.
The practical importance was that agents still needed strict authority boundaries because a fast but incorrect infrastructure action could spread failure across hybrid systems. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Southeast Asia Needed Data and Infrastructure First
Governed information, persistent memory and orchestration were prerequisites for agents that operated across enterprise systems. Without them, deployments would remain isolated assistants with limited agency.
The practical importance was that regional strategies also needed local language support, cross-border data controls and skills that allowed teams to inspect vendor-dependent orchestration. A credible evaluation would connect the claim to representative outcomes, record exceptions and human intervention, and identify who could act when conditions changed. It should compare different workloads or user groups and document the resources required to sustain the result, preventing one headline metric from substituting for operational evidence. Teams would also need a baseline, a review schedule and a defined threshold for pausing or changing the system. Those controls would make progress measurable while preserving room to respond when costs, risks or user needs diverged from the original assumptions.
Across the source report, technical capability was only one part of the system. Skills, access, integration, governance, cost and affected people could each determine whether the promised value survived outside a demonstration.
The durable lesson is to preserve traceability and choice. Organisations should document assumptions, keep consequential outputs reviewable, monitor downstream effects and retain a realistic exit route. Those controls allow ambitious AI work to improve without making early momentum an irreversible commitment.
