Agentic AI is already operating inside Singapore organisations, but the governance structures, job definitions, and workforce skills needed to manage it responsibly are failing to keep pace with deployment, according to new research from NTUC LearningHub. The Industry Insights Report on Agentic AI and Organisational Transformation, which surveyed 200 business leaders, finds that 15 per cent of respondents strongly agree that tasks within their organisations are already supported by AI agents — particularly in data analysis and reporting (46%), compliance (30%), customer support (30%), IT and cybersecurity (29%), and marketing and sales enablement (28%).

Job Scopes and Accountability Not Keeping Pace

Despite active deployment, organisational structures have not been updated to reflect AI agent involvement. Only 27 per cent of business leaders say their job scopes accurately reflect the involvement of AI agents, while 42 per cent report that AI responsibilities are under-represented and 25 per cent say job scopes have not been updated at all. The accountability gap is equally pronounced: more than a quarter of business leaders are not confident in their ability to supervise AI agents, and 24 per cent say accountability is unclear when AI agents are involved in decisions.

The report also identifies challenges in measuring AI agent performance, with fairness and bias cited by 39 per cent of respondents as the most difficult area to quantify, followed by accountability attribution (36%). The emerging roles most in demand as a result of agentic AI are data privacy and protection (46%), AI operations and performance monitoring (45%), and AI risk and compliance (43%).

Workforce Readiness Trailing Adoption

The skills gap is the most significant workforce challenge identified, with 44 per cent of business leaders citing a lack of skills required for AI-augmented roles. This is followed by difficulty measuring individual and team performance fairly in AI-integrated environments (43%) and limited access to relevant upskilling opportunities (42%). The top emerging competency cited is critical thinking when working with agentic AI-generated outputs (38%), followed by change management and agentic AI adoption leadership (37%).

Despite the push towards greater AI autonomy, 37 per cent of business leaders say their most important criterion for trusting AI agents is ensuring a human remains in the loop to review and approve recommendations before action is taken.

“AI agents are already embedded in daily operations, but many organisations are scaling faster than what their workforce and control structures can manage. As AI systems take on more autonomous decision-making, gaps in supervision, accountability and governance become real risks if not thought through and properly designed.” — Amos Tan, Assistant Chief Executive and Chief Core Skills Officer, NTUC LearningHub