AI and banks: Unlocking opportunity without creating risk

Increasingly, artificial intelligence (AI) is being woven into the fabric of business, becoming an integral part of decision-making and problem-solving for companies of all sizes.

The maturity of AI and its transformational promises couldn’t have come at a better time. The threats of online fraud and identity theft have become incredibly complex. Widespread digital adoption has made massive amounts of data available to criminals. Whether they’re stealing personally identifiable information through hacking and data breaches or targeting individuals through phishing and social engineering, criminals are continuously adapting their methods of attack.

As financial services providers digitize onboarding and lifecycle programs to better meet customer demands, the door to fraud has opened even wider. At the same time, consumer expectations are high, and competition is fierce. Approving more legitimate customers without friction while detecting fraud and maintaining compliance in a quickly changing digital landscape presents a sizable challenge that AI can help solve.

With its ability to quickly scrutinize vast volumes of data, AI is a boon for financial services organizations dealing with millions of daily transactions. AI automates threat discovery, accelerating the detection of patterns of suspicious activity for faster, enhanced decision-making. Yet, incorporating AI into the identity verification process can’t be taken lightly. AI can play an important role in identity verification and fraud detection, but its lack of transparency introduces serious risks. Fortunately, there’s a solution, but it’s also essential to understand the risks of relying solely on AI.

AI’s inherent weak spots

The inability of AI, and the machine learning models that power it, to provide visibility into the identity verification reasoning and decisioning process creates several obstacles. Without transparency, it’s impossible to explain to regulators why a decision was made or produce an auditable trail showing that onboarding policies were followed.

If a business cannot demonstrate how an outcome was reached, its compliance department is at risk for potential regulatory repercussions or class action lawsuits. Opaque AI also makes it difficult to prevent bias or confirm ethical AI decision-making. AI systems are trained on massive datasets that inform the decisions they make. When the data, algorithms or decision-making processes used in training an AI system are biased or under-representative, it can lead to unfair or discriminatory outcomes.

The machine learning models that power AI don’t have a moral bias; they simply do what they’re asked to do based on the information they’ve been given. Feedback from AI is essential for deeper data analysis and fine-tuning fraud detection models. Determining when and why an erroneous decision occurred is the only way to teach the machine how to spot similar issues in the future. Improving the learning model through constant input and data refinement, which includes training the system to recognize new fraud vectors that AI missed, is a critical step.

How can financial services organizations address AI’s weaknesses and tap into its power without increasing risk?

Humans and AI for the win

Layering AI with human expertise and intelligent verification technology makes it invaluable.

The benefits of human oversight alone are vast. Consider one of AI’s greatest blind spots—the inability to detect new forms of fraud. Because AI systems analyze data patterns, they assume future activity will follow those same patterns. This opens up the possibility of successful unsophisticated attacks simply because the system has not yet seen them.

Machines are great at detecting trends that have already been identified as suspicious but fail at novelty. A trained fraud analyst will catch novel threats missed by AI systems. AI, in general, is not error-proof. When it comes to a process as important as identity verification, one mistake can open the door to fraud, lead to noncompliance or damage customer relationships.

Human fraud analysts bring oversight, ethical decision-making and continuous improvement to AI. They also have contextual understanding and the ability to evaluate situations based on intuition and experience. If an AI system rejects a legitimate ID, a fraud analyst can jump in, determine how the error occurred and teach the computer how to spot similar issues in the future. This continuous feedback improves machine learning models through constant input and refinement.

On the other hand, an AI system without oversight will assume uncorrected bad behavior is accurate and will continue making the same decisions, thereby exacerbating the problem. This type of failure can draw scrutiny from regulators and result in fines and sanctions. It can also result in reputational harm, which, in turn, can cause financial losses.

Thinking strategically about AI

AI isn’t just being used for good. Criminals are using AI to create new forms of fraud, target victims and evade detection. Thus, financial services providers must think strategically about how they tap into AI to fortify themselves against threats without creating new vulnerabilities.

AI has the power to enhance or destroy a business, and successfully leveraging it for identity verification requires a complete ecosystem of fraud prevention. Multiple layers of intelligence are essential, from automation through AI for rapid fraud detection to fraud analysts.

Fraud analysts provide oversight and closed-loop transparency for continuous improvement and optimization. Through partnerships with software developers, they feed fraud intelligence into product innovation roadmaps, ultimately empowering users to prevent more fraud.

Fraud analysts are also invaluable in providing firsthand, expert insight into the fraud they’re seeing in the marketplace and best practices for preventing it.

As part of a multilayered approach, fraud analysts monitor transactional activity across a vast consortium network of customers, giving users a cross-industry view of fraud and fraud specific to their business so they can make more informed fraud-prevention decisions.

Knowledge is power, and understanding AI’s many facets and associated risks is the first step to success. With human oversight as part of a multilayered ecosystem, financial services providers can tap into the full potential of AI for identity verification and fraud prevention.

Heidi Hunter is Chief Product Officer at IDology

A version of this article first appeared in the BAI December Executive Report: “2024 Banking Outlook.”