Online fraud is spiraling, costing businesses tens of millions each year, and financial institutions are waking up to the reality that fraud is no longer just a compliance issue or a customer service trade-off. It is a core business risk that demands a strategic response.

Two new pieces of research, from fraud prevention companies Ravelin and Feedzai, paint a stark picture. According to Ravelin’s Global Fraud Trends 2025 report, online merchants lose an average of $10.6 million annually to fraud. But the real story is not just about volume. It is about inaction. The report highlights internal tensions where fraud teams want to act decisively, but leadership hesitates out of fear of damaging customer experience.

In parallel, Feedzai’s research team has released a pioneering AI framework, OpenL2D, alongside the FiFAR dataset, offering the financial services industry new tools to evaluate human-AI collaboration in fraud decision-making. Published in Nature Scientific Data, the research provides critical insights into how fraud detection systems can better defer to human judgment when the stakes are high.

These two contributions, one focused on frontline commercial losses, the other on long-term infrastructure and AI governance, together highlight the need for a smarter, more unified approach to fraud prevention in financial services and fintech.

Ravelin’s survey, which captured insights from over 1,400 fraud and payments professionals across 10 countries, found that 76 percent of businesses feel pressured to approve refunds even when they suspect abuse. This hesitation is most acute in sectors like travel and retail but the same tension is now spilling into financial services, where seamless digital experiences are seen as a competitive advantage and friction is feared like churn.

“Too many businesses are happy to dismiss fraud as a cost of doing business,” said Martin Sweeney, CEO of Ravelin. “Downplaying fraud to protect the customer experience is a false dichotomy. With the right intelligence, firms can distinguish fraudsters from legitimate customers and still deliver great digital journeys.”

Sweeney is especially critical of the idea that better fraud controls necessarily mean degraded user experience. “It is one of the biggest internal misconceptions,” he added. “Good fraud prevention actually helps good customers by removing unnecessary obstacles, while stopping bad actors in the background.”

The speed and sophistication of fraud tactics are accelerating as cybercriminals begin to harness generative AI. Deepfake identities, synthetic customers, and increasingly convincing phishing attacks are now targeting banks and fintech platforms. Refund abuse, once a minor problem, has evolved into a systemic vulnerability.

According to Jas Anand, Senior Fraud Executive at Feedzai, the arms race is well underway.

“Fraudsters increasingly harness generative AI to execute sophisticated scams,” said Anand. “The future lies in a smarter blend of automation and focused human oversight. AI and ML should be the first line of defense, but human analysts are still essential to interpret ambiguous cases and ensure ethical responses.”

Feedzai’s OpenL2D framework and FiFAR dataset aim to address exactly this challenge. By simulating expert decision-making across 30,000 real-world fraud cases, the framework allows financial institutions to evaluate when and how AI systems should defer to human judgment, critical in high-stakes decisions such as loan approvals, payment blocks, and identity verification.

The research shows that AI performance varies significantly depending on the capacity and diversity of available human experts. This has major implications for banks and fintechs looking to scale fraud operations while maintaining compliance and fairness.

“AI should identify patterns and anomalies at scale,” Anand said. “But people must provide the context. This is not about replacing fraud teams. It’s about making them more effective by giving them better tools and more time to focus on strategy.”

Beyond the technology itself, both companies point to a more systemic issue inside many financial institutions: data fragmentation. Fraud prevention teams are often working in isolation from product, operations, and customer support, which limits their ability to see the full picture of a user’s behavior.

“Unlocking the full potential of your data is one of the most overlooked changes a firm can make,” said Sweeney. “Your data tells you everything you need to know, who to trust, when to issue a refund, whether a transaction is legitimate. But many organizations don’t have the infrastructure to act on that insight.”

Anand agrees, arguing that enhanced, real-time data integration is a force multiplier. “Many institutions still focus on analyzing individual transactions rather than the broader behavioral context. This narrow lens results in higher false positives and missed threats.”

When browsing history, payment patterns, device IDs, customer support tickets, and risk scores are analyzed together, the decision-making becomes more accurate and more responsive. This unified approach reduces both fraud losses and unnecessary friction for trusted users.

Ultimately, both experts converge on one key point: fraud prevention must become a cross-functional, strategic capability, especially in financial services, where trust is currency and user experience drives retention.

“To shift internal mindsets, fraud needs to be treated as an investment in trust and long-term growth,” said Anand. “Show ROI using internal case studies. Promote awareness beyond the risk team. Integrate fraud controls into product launches, marketing campaigns, and operational workflows.”

Financial institutions should also consider contributing to collective intelligence networks, where anonymized threat signals are shared across the industry. Fraud rings rarely target a single company. Real-time collaboration can prevent multi-platform abuse and provide earlier warnings.

1. Reposition fraud as a value driver, not a cost center
Link fraud prevention to brand trust, user retention, and product innovation

2. Break down data silos
Unify e-commerce, support, risk, and product data for real-time, contextual decision-making

3. Adopt hybrid intelligence systems
Use AI to scale pattern recognition, but reserve human analysts for nuance and escalation

4. Upgrade refund abuse detection
Track behavioral patterns over time, flag high-risk users, and deploy smart verification methods

5. Invest in cross-functional education
Make fraud KPIs visible across leadership. Align product, compliance, and customer experience teams

6. Leverage shared threat networks
Join anonymized data-sharing initiatives to detect coordinated fraud earlier

Fraud is no longer just a technical problem for the back office. It is a strategic maturity test for financial institutions, especially as they race to modernize infrastructure and deliver seamless digital experiences. As AI reshapes both the threat landscape and the toolkit available, the firms that succeed will be those that recognize fraud prevention as a cornerstone of customer trust and revenue protection, not a trade-off.

For more like this on Forbes, check out AI’s Growing Role In Financial Security And Fraud Prevention and Risk-Based Authentication: The Future Of Secure Digital Access.