Background of Anti-Money Laundering 

Financial crime is evolving faster than ever, and money laundering has become more sophisticated, faster, and harder to detect using traditional, human-only oversight.

As electronic banking continues to grow, so do the opportunities for financial criminals. This is especially true for smaller banks still relying on legacy Anti-Money Laundering practices, where manual monitoring struggles to keep pace with the volume and speed of digital transactions.

The impact on Malaysia is significant.

In 2024, money laundering accounted for 5.04% of Malaysia's GDP, translating to more than USD 20 billion in losses. At the same time, the yearly Costs of Compliance (CoC) reached USD 1.95 billion. Studies by Napier AI suggest that AI-driven AML solutions could reduce these costs by nearly 50%, representing potential savings of USD 0.96 billion.

While official figures from the Malaysian government are limited, the scale of the problem is further illustrated by data from the Penang Institute, which estimates USD 12.8 billion in losses from online scams alone.

Taken together, the complexity of modern financial crime, the size of the financial losses, and the rapid growth of Malaysia's digital economy clearly point to one conclusion: traditional AML approaches are no longer sufficient.

To keep up, financial institutions need a faster, more accurate, and highly scalable approach to Anti-Money Laundering, one that can evolve as quickly as the threats it is designed to stop.

Opportunities for Agentic Automation in AMLA 

Addressing the growing complexity of financial crime requires more than incremental improvements. The real opportunity lies in Agentic Automation.

By combining advanced machine learning, Large Language Models (LLMs), and autonomous decision-making agents, financial institutions can move beyond static, rules-based transaction screening. Instead, they gain intelligent systems that actively reason, analyze, and report on financial transactions in real time.

These AI agents are able to interpret and retain policies and regulatory requirements far faster than human teams, while remaining fully auditable and subject to human oversight. The result is a more efficient AML process, with lower computational costs, fewer false positives and negatives, and significantly faster investigations.

Human oversight, supported by intelligent systems

Regulatory frameworks in Malaysia already support this model. Policies from the Central Bank of Malaysia place the responsibility for transaction flagging firmly with individual banks. When transactions breach internally defined thresholds, such as a high volume of small transactions within a short period, they must be monitored and, where necessary, reported to authorities.

While human judgement remains essential, Agentic Automation dramatically reduces the manual workload. AI agents can pre-screen transactions, prioritize risk, and surface only the most relevant cases for review. This allows compliance teams to focus on decision-making rather than data processing.

Proven efficiency gains

The adoption of AI in AML is growing, but there is still significant room for improvement. In a 2023 investigation by Bank Negara Malaysia, 21 out of 25 surveyed reporting institutions indicated they were running AI or machine learning projects. However, only 13 were using these technologies specifically for fraud or AML detection.

This gap represents a major opportunity.

AI-driven AML systems not only reduce operational costs but also dramatically shorten investigation times. According to Napier AI, AML work time can be reduced by up to 65% when using AI-powered platforms. In practice, this means only 30% of AML work requires human intervention, with 70% handled by automated systems.

For Malaysian financial institutions, these efficiencies translate directly into lower compliance costs, faster response times, and a more scalable approach to meeting regulatory obligations.

Use of Agentic Automation in AMLA 

Agentic Automation transforms the AML process by introducing intelligent AI agents that continuously screen, analyze, and assess financial transactions.

These agents first apply advanced machine learning models to evaluate transactions and estimate the likelihood of fraudulent activity. When a transaction exceeds a predefined risk threshold, it is automatically flagged and escalated to a human compliance officer for review. This ensures that critical decisions remain under human control, while routine analysis is handled at machine speed.

In parallel, Large Language Models (LLMs) enable AI agents to interact with human users and interpret large volumes of AML policies, regulatory requirements, and internal guidelines. By retaining and applying this knowledge consistently, AI agents support more accurate and standardized decision-making across institutions.

The impact is substantial. Agentic Automation significantly reduces the need for manual effort, lowers operational costs, and enables financial institutions to operate more efficiently. At a broader level, it contributes to a safer and more resilient digital financial ecosystem for Malaysia.

Balancing automation with data privacy

However, increased automation also introduces new challenges, particularly around data protection and privacy. Under Malaysia’s Personal Data Protection Act (PDPA), financial institutions are responsible for safeguarding sensitive customer information.

As large volumes of data are processed by AI systems, the risk of exposure to cyber threats increases. This makes robust security measures and strong governance frameworks essential. AI platforms must be supported by secure infrastructure, well-defined access controls, and compliant data governance, ensuring that sensitive information remains protected from misuse or unauthorized access.

By combining Agentic Automation, strong cybersecurity, and regulatory compliance, financial institutions can unlock the full potential of AI-driven AML while maintaining trust and protecting consumer data.

Figure – How Agentic Automation Works in AML

Conclusion 

As financial transactions increasingly shift to digital channels, Malaysia is facing substantial losses from money laundering, while the cybersecurity of these systems struggles to keep pace. In 2024 alone, losses exceeded USD 20 billion, underscoring the urgent need for more effective and scalable Anti-Money Laundering approaches.

Agentic Automation offers a clear path forward. High-speed, intelligent AI agents can model AMLA processes with far greater efficiency, significantly reducing manual effort and improving detection accuracy. At the same time, the Costs of Compliance could potentially be reduced by nearly 50%, freeing up resources while strengthening regulatory outcomes.

These gains position AI to play a central role in transaction monitoring and control, provided that auditability, transparency, and human oversight remain firmly in place. Automation must support human decision-making, not replace it.

Equally important is the protection of personal data. As sensitive financial information is processed at scale, strong data protection, governance, and cybersecurity measures are essential. Privacy risks associated with digital platforms must be addressed through clear legislation, robust supervisory frameworks, and accountable oversight bodies.

Moving forward, Malaysian reporting institutions, in close collaboration with AI solution providers and regulators, should prioritize the adoption of auditable, AI-driven AML systems as a core component of their financial infrastructure. By aligning innovation with governance and privacy safeguards, Malaysia can significantly improve the effectiveness of its AML efforts and take meaningful steps toward reducing the hidden costs of financial crime in its digital economy.