Financial services firms are pouring billions into artificial intelligence, yet most remain trapped in experimental phases rather than deploying AI at scale. The disconnect between enthusiasm and actual implementation reveals a deeper challenge: institutions don't lack the technology or the business case for AI. What they lack is confidence that AI systems can operate reliably within regulated environments where decisions must be explainable, auditable, and defensible.

What's Actually Holding Back AI in Finance?

A comprehensive survey of financial services leaders and technology vendors uncovered a striking pattern. When asked about barriers to AI adoption, nearly 60% of institutions in the Asia-Pacific region flagged model performance and reliability as their top concern, with close to 50% in the UK, Europe, and North America citing the same issue. Governance and control frameworks ranked as the second major barrier across all regions.

But here's what's revealing: when the same institutions were asked about return on investment (ROI) and internal expertise, these ranked at the bottom of their concerns. This suggests the business case for AI in finance is largely accepted, and the skills to implement it are more available than they once were. The real blocker, according to industry leaders, is trust.

"Model performance and reliability tops every region, and governance and control frameworks follow closely. These are barriers. Hallucinations in a KYC workflow or an unreliable model in transaction monitoring carry real regulatory and financial consequences," said Areg Nzsdejan, CEO of Cardamon.

Areg Nzsdejan, CEO of Cardamon

The concern about reliability isn't theoretical. In anti-money laundering (AML) systems, for example, models can produce plausible-looking answers from flawed risk assessments. These "type 3 errors," where a model treats a correlated data point as a causal one, can pass initial testing because the output looks correct, yet fail under regulatory scrutiny even when accuracy metrics appear strong.

Why Governance Is the Real Constraint?

Financial institutions have spent decades building risk frameworks for traditional operations. Artificial intelligence, particularly generative AI, introduces new capabilities and new behavior patterns that existing governance structures weren't designed to handle. According to recent data, 88% of organizations already report governance and security challenges when adopting AI, with 53% of AI agents exceeding their intended permissions and 47% of organizations experiencing an AI agent-related security incident.

This governance gap is what's keeping pilot projects from reaching production. The problem isn't that AI hallucinations or model drift are new concepts; financial institutions understand operational errors. The challenge is that AI introduces these familiar problems in unfamiliar packaging, requiring new oversight mechanisms and accountability structures.

"In a regulated environment, decisions need to be explainable, auditable and defensible. Institutions therefore need to know not only that an AI model performs well, but why it produces a particular outcome, how its performance is monitored and who remains accountable for the resulting decisions," explained Jean Voigt, head of AI at IMTF.

Jean Voigt, Head of AI at IMTF

How Financial Firms Are Moving From Pilots to Production

Despite these barriers, financial institutions are finding practical pathways forward. Rather than pursuing full automation, many are adopting hybrid systems where technology handles routine processing while employees focus on unusual cases, disputes, and higher-risk decisions. This approach maintains human judgment where it matters most while capturing efficiency gains from automation.

The transition from experimental AI to operational deployment requires attention to several critical areas:

• Data Quality and Governance: Financial organizations need robust systems for managing data permissions, ensuring auditability, and maintaining cybersecurity standards before deploying AI systems that influence payments, account access, or customer-facing decisions.
• Explainability and Transparency: Models must be able to explain their reasoning in ways that regulators and customers can understand. This is particularly important in compliance workflows, fraud detection, and customer onboarding processes.
• Escalation Procedures: Clear pathways for human review are essential. When an automated decision is incorrect, customers need to know how to escalate the issue and reach someone who can resolve it quickly.
• Hybrid Governance Frameworks: Different types of AI (deterministic rules, machine learning, network analytics, and generative AI) have different strengths and transparency levels. Combining them strategically allows institutions to apply the right technology to the right problem while maintaining appropriate oversight.

Financial institutions are also recognizing that emerging AI applications don't necessarily require abandoning established payment infrastructure or consumer safeguards. A European transaction demonstrated how an AI agent completed a purchase while operating within existing banking and authentication rules, illustrating that practical AI integration often works by building on systems customers and institutions already understand.

The Broader Shift in Financial Services Operations

Across the financial sector, conversations about AI are increasingly moving beyond theoretical capabilities and focusing on practical implementation. Banks, payment providers, brokers, and fintech companies are now exploring automation in customer onboarding, transaction monitoring, fraud detection, risk management, document processing, customer support, and payment workflows.

This operational focus reflects a maturation in how the industry thinks about AI. Rather than treating every new AI capability as a fundamental transformation, financial institutions are examining how technologies perform once they enter regulated environments. This includes scrutinizing the infrastructure behind new services, the companies implementing them, the regulatory frameworks governing their use, and the operational consequences for financial institutions.

"The real question is not whether AI is inherently too risky for financial services, but which type of AI is appropriate for a particular use case and which controls are needed around it," noted Jean Voigt.

Jean Voigt, Head of AI at IMTF

As artificial intelligence becomes more deeply embedded in financial services, topics such as explainability, cybersecurity, operational resilience, data governance, and human accountability are becoming increasingly connected to discussions about automation. Technology may change rapidly, but financial institutions will continue operating within an environment where reliability and trust remain essential.

The institutions that successfully move AI from pilots to production won't be those with the most advanced models. They'll be the ones that build governance structures first, maintain clear human accountability, and design systems where customers understand how decisions are made and can resolve problems when they arise.