Singapore’s financial industry is facing a reckoning. In June, the Monetary Authority of Singapore (MAS) imposed over $3.8 million in civil penalties against financial institutions and individuals for breaches related to anti-money laundering and market misconduct. Similarly, as regulators across Asia are becoming more aggressive and more aligned on tracking down compliance failures, it’s clear that the old compliance playbook will no longer cut it.
However, amid this heightened scrutiny, a new generation of artificial intelligence, or agentic AI, is emerging as a pivotal solution. For financial institutions across Southeast Asia wrestling with tougher regulatory scrutiny and pressures to digitalise, agentic AI offers real-time learning, decision-making and adaptability beyond rule-execution.
Unlike traditional AI that runs predefined models and automates routine processes, agentic AI can act with a degree of independence – scanning transactions and exposing anomalies, while adapting workflows dynamically, acting autonomously within boundaries. These systems do not just follow rules; they learn, decide, and take action, presenting a fundamentally different approach to managing compliance in real time.
According to our research conducted in collaboration with Chartis Research and launched last month, 6% of financial services firms have already implemented agentic AI, while 93% plan to do so within the next two years. Fraud detection emerged as the top use-case at 36%, followed by KYC maintenance at 19% and transaction monitoring at 16%.
This rapid uptake reflects the pressure from regulators like MAS, as well as internal demands to reduce costs, eliminate backlogs, and onboard clients seamlessly without compromising on risk controls.
But adoption comes with challenges. MAS has made it clear that outsourcing risk to a machine is not an excuse for compliance failures. Agentic AI decisions must be explainable, traceable, and accountable, and human oversight and strong governance must remain at the centre. If you get agentic AI wrong, you risk replacing one blind spot with another, which will not hold up well with boards and regulators.
See also: Scaling tokenisation in regulated finance to unlock real-world financial value in Singapore
Transparency is a non-negotiable. Every action, from clearing a false positive to escalating a flag, must be explainable and traceable. If your AI cannot articulate its logic in plain language, you will have a liability on your hands, not innovation.
Human governance matters just as much. Southeast Asia’s regulatory terrain and cultural complexity mean no agent should act alone. The most credible implementations today use a hybrid model, where agentic AI surfaces patterns and updates risk profiles continuously. However, compliance officers still need to make the final call. With enhanced decision-making capabilities from agentic AI, this human-in-the-loop model ensures that a balance is maintained between automation and accountability.
Singapore’s push to combat money laundering has accelerated this shift towards agentic AI. Last year’s billion-dollar laundering scandal has sparked MAS’s requirements for stronger customer due diligence, real-time monitoring and clarity on beneficial ownership. Agentic AI maps neatly onto these priorities by scoring customer risk in real time (without relying on outdated periodic reviews), analysing behavioural anomalies that rule-based systems may miss, and evolving new typologies.
See also: Driving the lowest TCO for data-intensive technologies and applications in Singapore
In addition, Singapore’s collaborative compliance model, which encourages financial institutions to share typologies and insights, would well be amplified with agentic AI trained on communal, anonymised risk data. This shift could transform compliance, significantly improving the sector’s collective ability to identify and respond to emerging threats, moving away from defending single institutions to defending the entire ecosystem.
Still, none of this exists outside regulatory frameworks that ensure transparency, fairness, and privacy. Stricter data protection regimes in Singapore, Malaysia, and Indonesia, plus emerging regional AI governance standards demand that financial institutions invest in explainable AI models to enforce data lineage controls and deploy governance mechanisms that align with both local and cross-border expectations.
The stakes are high, but the potential payoff is real. Our study tells us that over a quarter of respondents expect more than US$4 million in compliance savings annually from agentic AI deployment, driven by reduced manual workloads, faster decisions and fewer breaches. But the real value lies in institutional credibility. At a time where one compliance mishap can erode public trust and wipe out market cap, agentic AI can give Southeast Asia’s institutions a chance to act faster, see further and respond smarter.
The message from Singapore’s regulators is clear: the era of passive compliance is over. Agentic AI is beyond another digitalisation tool; it’s a litmus test for whether financial institutions in the region can reconceptualise compliance as an intelligent defence system.
Cengiz Kiamil is the managing director for Apac at Fenergo
