AI is transforming the financial sector, opening doors to new opportunities while presenting its own set of challenges. From adapting legacy systems to fostering a culture of innovation, the journey toward building sustainable AI-driven management systems in finance requires a strategic, multifaceted approach. Here, I’ll walk through 9 of the essential factors and share my insights into what it takes to get AI right in financial crime compliance.
1. Overcoming legacy system challenges
The biggest initial hurdle that many traditional institutions face is their own legacy systems. A staggering number of AI projects struggle to progress beyond pilot stages, largely due to infrastructural incompatibilities. Studies show that banks and insurers allocate over 70% of their IT budgets to maintaining these older systems, leaving little room for transformation.
As I often emphasise, the tech interface is foundational—if your existing systems can't support AI, you’re limiting its potential. Strategic updates to legacy systems are essential for AI to be more than just a theoretical asset.
2. Data quality and creating a 'Single Source of Truth'
Data is the lifeblood of AI. But, too often, firms have data lakes that resemble “sewage pits” rather than clean, structured repositories. This poor data quality comes at a massive cost, estimated at nearly $13 million per year for businesses. Without a single, reliable source of truth, AI can generate inaccurate outputs, creating regulatory and operational risks. Cleaning up data lakes and establishing a trustworthy data source is more than just a technical necessity; it’s foundational to actionable and reliable AI-driven insights.
Read more on AI for AML: Battling bad data to make data science magic
3. Addressing the skills gap and embracing diversity
Building an AI-ready team is about broadening the talent pool beyond the traditional financial and legal experts. Research reveals that only 26% of organisations believe their current workforce has the skills needed to handle AI effectively. This gap in skills isn’t just about tech know-how—it’s about having diverse perspectives that foster innovation and adaptability. Expanding teams to include data scientists and engineers, and investing in upskilling existing staff is essential to making the most of AI. To leverage AI fully, we need to support our teams with new skills and viewpoints.
4. Cultivating a culture of innovation and psychological safety
AI development is inherently experimental, and success requires a culture that allows for calculated risks. Psychological safety—the ability to make mistakes and learn without fear of repercussion—is a must for fostering innovation. Research shows that teams with high psychological safety are 76% more likely to innovate and 50% more productive. In our highly regulated industry, this mindset can be difficult to establish, but regulators must also evolve to allow safe experimentation within compliance frameworks. Innovation stems from making mistakes, and without that freedom, we stifle our ability to progress.
5. Balancing compliance and innovation
Balancing regulatory compliance with innovation has always been a tightrope for financial institutions. Historically, we’ve applied quick fixes to meet compliance, often at the cost of long-term growth. My view is that every regulatory change or risk development should trigger a reassessment of processes, pushing us toward a proactive approach to compliance. This is a top concern for financial institutions today, with 63% citing the balance between innovation and compliance as one of their biggest challenges. Partnering with regulators through public-private collaboration is one path forward, creating a shared vision that aligns compliance with progressive growth.
6. Avoiding common pitfalls in AI adoption
AI is no silver bullet, and the excitement it generates often leads firms to implement it without assessing internal needs. Too often, institutions treat AI as an off-the-shelf solution for singular issues, like Know Your Customer (KYC) or fraud detection, rather than a toolkit for holistic risk management. The fact that banks spent over $9 billion on AI last year—with 25% of those investments at risk due to poor alignment—tells us that we must tailor AI to fit organisational goals before investing. Understanding how AI fits into our unique structures and risk profiles is critical.
7. Collaboration as a foundation for risk management and trust
A collaborative approach is vital for effective AI-driven risk management. Open, closed-door discussions enable us to share challenges and solutions, building trust and strengthening the industry's approach to risk. Running the race alone isn’t sustainable. Collaboration across the sector, particularly on AI and risk management, sets a foundation for mutual success, and it’s a path we need to embrace more as the challenges we face continue to grow in complexity.
8. Re-envisioning training and shifting mindsets
For sustainable AI integration, training must go beyond technical skills. It’s about fostering an open mindset that champions innovation and adaptation. Traditional approaches can inhibit progress, so training needs to instill a proactive approach. At Napier AI, for instance, our team is trained not only in product-specific knowledge but also in financial crime trends and regulatory updates. This industry-wide knowledge base enables us to apply AI meaningfully. According to recent studies, over half of all employees will need reskilling or upskilling in the coming years, and it’s imperative we stay ahead.
9. Measuring AML effectiveness with AI
AI has already demonstrated its potential in anti-money laundering (AML) efforts. For instance, HSBC’s partnership with Google resulted in a tenfold increase in AML detection efficiency. Effective measurement through robust Management Information (MI) reporting is key for refining AI tools. I’m particularly interested in the future of AI-powered behavioral analytics, which promises to reveal deeper crime trends, victim profiles, and geographic hotspots. This capability could greatly enhance our ability to detect and prevent financial crimes before they escalate.
Building sustainable AI systems in financial services is a journey that requires thoughtful planning, ongoing collaboration, and a willingness to adapt. By prioritising data integrity, fostering a psychologically safe environment, and bridging the skills gap, we can unlock AI’s full potential while remaining compliant and resilient. As AI technology evolves, these guiding principles will help the financial industry innovate responsibly and effectively, paving the way for a more adaptive and forward-thinking approach to risk management.
Explore the first in-depth ranking of AI's impact on anti-money laundering in the Napier AI / AML Index.
Photo by Aakash Dhage on Unsplash
