By Jason Shane

Three days. Over 1,400 compliance practitioners. One undeniable truth: Malaysia’s banking sector is at an inflection point. At the 15th International Conference on Financial Crime and Counter Terrorism Financing (IFCTF) themed, The Future is Now: Tech-driven Compliance in the Fight Against Financial Crime, the conversation wasn’t just about compliance challenges. It was about reimagining them as competitive opportunities.

The conference brought together global and regional leaders to explore artificial-intelligence (AI)-enabled fraud detection, machine-learning-driven compliance, virtual assets and digital threats, terrorism and proliferation financing, and anti-bribery and anti-money laundering (AML) frameworks. What emerged from these sessions were clear: Malaysia faces a perfect storm of complexities: rising transaction volumes, sophisticated money laundering schemes, and digital payment innovations – all demanding compliance vigilance at scale. Yet, the traditional compliance model, heavily manual, resource intensive, and reactive, is cracking under these pressures.

But here’s what excited me most at IFCTF: the realisation across our banking community that this isn’t a crisis of capacity. It’s an opportunity for transformation. The answer isn’t hiring more compliance staff or implementing AI as a wholesale replacement for human judgement. It’s something far more powerful: the 50/50 model, a blueprint that lets technology and human expertise do what they do best, together.

The Trap of Extremes

For decades, Malaysian financial institutions have operated under a largely manual compliance regime. Investigators review alerts, dig through transaction histories, cross-reference multiple systems, and make judgement calls on suspicious transaction reports (STRs). It’s thorough. It’s grounded in experience. It’s also becoming hard to scale and meet evolving compliance and threat challenges.

Transaction volumes continue to explode. Compliance teams are consumed by routine tasks, reacting to alerts, reconciling reports, and meeting mounting documentation demands. This leaves little capacity for proactive financial crime prevention. These constraints not only impede efficiency but also increase regulatory risk, operational cost, and reputational exposure. As Malaysia’s regulatory expectations grow, the traditional model is straining to keep up.

Meanwhile, the swinging pendulum suggests a tempting alternative: full automation. Let AI handle everything, some argue. Remove the human bottleneck entirely.

But full automation is a false promise, especially in financial crime prevention. Deciding whether to freeze a customer’s account, filing a suspicious activity report (SAR) that carries regulatory weight, or interpreting ambiguous transactions considering evolving regulations. These require judgement, context, and accountability. They demand human wisdom. Regulations in Malaysia increasingly demand this human accountability, holding individuals personally liable for institutional failures. No algorithm should bear that burden alone.

The real power lies between these extremes.

The 50/50 Model: Balance as Competitive Advantage

The 50/50 model allocates half of compliance operations to AI-driven automation and the other half to human expertise. It’s a framework, not a rigid formula. The precise split varies by financial institution, but the principle is profound: let AI handle volume, consistency, and speed; and let humans handle nuance, judgement, and risk oversight.

In practice, here’s how it works:

> The AI 50%: Detecting anomalies in real-time, enriching alerts with data from multiple systems, prioritising cases by risk, pre-filling reports with context, and executing predictive models that identify suspicious patterns before they balloon into full investigations. AI agents, agentic AI, orchestrate these workflows with the speed and consistency that no human team can match. A transaction screening process that took hours now happens in a few minutes. Sanctions lists update automatically and customers are re-screened instantly. False positives that previously flooded investigators with noise are eliminated upfront. What emerges is a curated set of genuine threats deserving expert attention.

> The Human 50%: Making the judgement calls on complex cases. Interpreting regulatory ambiguity. Challenging AI recommendations when context demands it. Filing SARs with confidence, knowing they’re backed by human reasoning. Adapting policies as regulations evolve, as they constantly do in Malaysia. Providing oversight that satisfies regulators and auditors alike. Humans bring accountability, ethical reasoning, and adaptive judgement. All of which are irreplaceable in financial crime prevention.

Why Malaysia Needs This Now

Malaysia’s regulatory landscape is particularly well-suited to the 50/50 model. Bank Negara Malaysia’s (BNM) heightened focus on real-time monitoring, enhanced due diligence, and ownership transparency creates immense volume. The Anti-Money Laundering, Anti-Terrorism Financing and Proceeds of Unlawful Activities Act 2025 amendments demand more rigorous sanctions screening and STR reporting expectations. Financial institutions that don’t automate the bulk of routine tasks will be buried in backlog and high costs of resources. Those that go to automation without human oversight risk regulatory censure and reputational damage.

The 50/50 model strikes the balance Malaysia’s regulators and industry leaders emphasise. During IFCTF sessions and in recent industry discussions, a consistent theme emerged: financial institutions must adopt AI responsibly, with strong human oversight and governance. Malaysia’s first AI Governance Framework for financial services, developed by the AICB’s Chief Risk Officers’ Forum in collaboration with BNM, emphasises exactly this: responsible adoption, human oversight, and the preservation of public trust.

Consider a real case: a major financial institution reduced manual processes of assessing transactions flagged as sanctions hits, reducing investigation time by 10x, saving 90 minutes per case. How? By letting AI agents handle triage, data enrichment, and pattern matching. The result? 98% case agreement between AI-assisted investigations and human analysts, plus a 99% reduction in false positives. That’s not replacing people; that’s amplifying them.

The Workforce Evolution

Implementing the 50/50 model isn’t just about technology; it’s about people. Compliance teams shift from routine investigation to strategic risk analysis. Investigators evolve into ‘case reviewers’ and ‘judgement experts’. New roles emerge: AI auditors who validate agent decisions, model risk managers who oversee algorithmic fairness, and compliance strategists who interpret regulatory signals.

This transition requires investment in training and culture. Staff need to understand how AI works, trust its recommendations, and know when to override them. But the payoff is significant: reduced burnout, higher-value work, and a compliance function that attracts talent rather than losing it to competitors.

For Malaysian financial institutions, this is particularly important. The compliance talent market in the region is tight. The AICB’s 2025 Workforce Survey found that more than 40,000 banking employees are expected to see their roles evolve due to automation and technological augmentation. Offering roles that blend human expertise with AI tools, roles that are intellectually stimulating rather than repetitive, becomes a competitive advantage in recruiting and retaining the best practitioners.

The ROI: Beyond Cost Savings

Let’s talk numbers. Many compliance leaders approach AI with a cost-reduction lens: fewer investigators needed, lower operational costs. That’s part of the story, but it undersells the real value.

Yes, the 50/50 model reduces costs. By eliminating manual false positive triage and automating routine investigations, financial institutions see dramatic efficiency gains. But the bigger story is risk reduction and strategic value:

+ Better detection

AI doesn’t tire. It doesn’t miss patterns because an investigator was overwhelmed that day. It catches subtle links between transactions that span months and multiple customer relationships.

+ Faster response

BNM expects financial institutions to respond to suspicious activity quickly. Shorter investigation times mean faster SARs, which means better regulatory standing and reduced reputational risk.

+ Regulatory confidence

 Auditors and regulators see AI-assisted investigations with full explainability and robust human oversight. This builds confidence in a financial institution’s compliance posture.

+ Competitive advantage

In Malaysia’s increasingly crowded financial services market, financial institutions that are known for robust, proactive compliance attract higher-quality customers and business partners.

The Path Forward: Incremental, Not Revolutionary

I emphasise this to every financial institution: implementing the 50/50 model doesn’t require a full overhaul of your compliance function. It’s incremental. Start with high-volume, manual processes. Deploy AI overlays on your existing systems. Measure results. Build internal expertise and confidence. Then expand.

The key is to start now. Malaysia’s regulatory environment moves quick. Financial institutions that begin their AI journey today will have months of learning and optimisation under their belt when the next regulatory wave hits. Those that wait will be playing catch-up.

A Call to Action: Embrace the Balance

The IFCTF conference underscored a vital insight: Malaysia’s compliance community understands the magnitude of the challenge and the urgency of transformation. But understanding and acting are different things. Many financial institutions are still locked in the ‘all-human’ or ‘all-AI’ debate, paralysed by choices which aren’t binary.

The 50/50 model isn’t a theoretical ideal. It’s achievable regardless of your starting point. And it’s essential to thrive in Malaysia’s compliance landscape. The question isn’t whether your financial institution has the capacity to implement AI-driven compliance. It’s whether you have the capacity not to.

Jason Shane is a seasoned technology executive with more than 30 years of leadership experience across global financial institutions. He brings deep expertise in IT strategy, product innovation, and transformation within compliance, risk, and financial crime prevention domains. At SymphonyAI, he leads global product strategy and innovation, where he partners with financial institutions to transform compliance from a cost centre into a competitive advantage through AI and human expertise.

SymphonyAI builds vertical AI applications that help enterprises tackle their most complex, high-value challenges — like stopping financial crime, improving store performance, and boosting manufacturing efficiency. Trusted by more than 2,000 enterprise customers in 26 countries, including 200 of the top financial institutions, top 25 CPGs, and many of the world’s largest grocers and industrial manufacturers.