Trust has always been the foundation of the credit union movement. Today, that trust is facing one of its greatest tests.
Artificial intelligence is making fraud more convincing, more scalable and more difficult to detect. As AI becomes more embedded in commerce, credit unions face a difficult balancing act—embracing innovation while protecting members from evolving threats.
In a recent PaymentsJournal podcast, Karen Postma, Senior Vice President of Risk Solutions at Velera and Suzanne Sando, Lead Fraud Management Analyst at Javelin Strategy and Research, discussed how AI-driven fraud, deepfakes and the emergence of agentic commerce could jeopardize the critical bond between members and credit unions.
Not only has this made it imperative for institutions to implement robust technological safeguards, it has also highlighted the importance of effective education and communication in maintaining strong member relationships.
Questioning the Source of Truth
Criminals have forged ahead in the AI arms race, largely because they aren’t constrained by compliance, operational or ethical obligations. As concerning as this is for financial institutions, consumers have also become aware of the risks as well.
Many have seen first hand how AI has made fraudulent communications more convincing through sophisticated scam texts, phishing emails and other impersonation attempts. And they understand these tactics are just the tip of the iceberg.
“I don’t think this is an overreaction,” Sando said. “We have information overload when it comes to AI. From a consumer perspective, you’re hearing about it in the news all the time—the good, the bad and the ugly. You’re hearing about the ethical debates and the effects on the environment. We can’t get away from it, and that may play into some of the fears that consumers have with AI.”
Consumer concerns have grown as deepfakes have demonstrated how convincingly AI can mimic a person’s voice or likeness in audio and video. These tools can be used to deceive friends and family, authorize fraudulent transactions or facilitate blackmail and other scams.
However, deepfakes are also indicative of the double-edged nature of AI. While consumers worry about becoming victims of these attacks, technology is also lowering the barrier for individuals to commit first-party fraud.
“That happens all the time, and then they’re disputing it or claiming some type of loss when in actuality that wasn’t the case,” Postma said. “Unfortunately, the accessibility of innovation in AI has made us question everything more, versus trusting in the consumer.”
“That’s a weird position to be in. The member has always been our best source of truth, and unfortunately, I don’t know that we’re in that environment any longer,” she said.
The Emergence of Agentic Commerce
This dual nature of technology has also raised concerns about the emergence of agentic commerce, where consumers entrust AI agents to act as autonomous personal shoppers. Consumers already use AI to compare prices, research products and find the right item.
However, fully autonomous agentic commerce is an entirely different proposition.
“You’re trusting an agent to understand your intent, and you’re also trusting that agent to be able to execute in the way that you would want them to,” Postma said. “Such things as understanding intent from the initiation of the prompt, all the way through the transaction, all the way to a potential dispute process—and making sure that transparency is there. That’s going to be a critical avenue where the industry is going to have to adapt and figure out how to share that information.”
Beyond the foundational challenges of building these systems, fraud prevention becomes more complex. Rather than authenticating a transaction, financial institutions will also need to verify the customer’s intent and ensure the AI agent is acting according to the customer’s wishes.
Introducing a third party into the traditional transaction model also creates new avenues for fraud. Both the agent and the customer will require authentication, potentially necessitating the development of Know Your Agent policies that mirror Know Your Customer requirements.
This could include building profiles for AI agents that document aspects like behavioral patterns, permissions, and preferred merchants. These profiles could help identify when an agent has been compromised or manipulated.
Alongside these infrastructure changes, the returns process—already a common target for fraud—will also need to evolve.
“The number of disputes increases drastically—I’m using the word ‘dispute’ intentionally and not ‘fraud’—because in those cases, consumers forget that they empowered an agent to go purchase this for them,” Postma said. “They don’t recognize the transaction because they empowered that agent a couple weeks ago.”
“Those kinds of things happen in that fully autonomous world, and we don’t necessarily have the infrastructure to mitigate those disputes at the moment,” she said.
High Stakes and Tight Budgets
As credit unions work to defend against AI-driven fraud while preparing for agentic commerce, many are developing strategies to modernize their fraud prevention programs. However, because these investments have far-reaching implications, institutions are often challenged to determine how best to allocate limited fraud prevention resources.
“Fraud is always at the bottom of budget prioritization, whether we like it or not,” Sando said. “It’s either compliance or regulation updates that take priority; it’s some enhancement that will benefit customers; it’s a new product that’s going to bring a new revenue. At the end of the day, it’s, ‘Whatever we have leftover, we’ll throw at the fraud problem.’ That’s never going to be enough if fraud doesn’t get the prioritization that it not only deserves, but it needs.”
Given the high stakes and tight budgets, partnerships can provide credit unions and community banks with an effective way to balance speed to market with the need to adapt to a rapidly changing environment.
“There are starting to be a number of providers in the marketplace, Velera being one of them, that allow for what we call an orchestration layer,” Postma said. “It allows the financial institution to integrate once on the back end, and it is powered by a number of different solutions. With such things as orchestrators to be able to leverage some of that technology, it becomes much more strategic and intentional around how we prevent fraud.”
Education, Transparency and Trust
Technology alone is not enough. Equally important are educating members and fostering transparency.
Too often, banks and credit unions rely on generic fraud education that fails to resonate with customers.
“You need to have real-world and contextual examples, things that are specific to someone’s demographic. Like saying: ‘We’re seeing a higher concentration of elders being targeted by a certain scam or people who are Gen Z or who use digital wallets,’” Sando said. “If you’re putting it into context, it helps put the threat into a space that doesn’t feel scary and insurmountable, but it does feel real.”
As scam communications have become more convincing, many customers have begun ignoring all digital messages—including legitimate fraud alerts from their financial institution.
This makes proactive, ongoing education essential. Credit unions should help members recognize common fraud tactics and understand how the institution communicates with them. Given the sophistication of today’s fraud landscape, that education must extend far beyond a handful of documents buried within an online banking website.
“It’s education internally, it’s education to the consumer, and from a credit union perspective and the role that we play in communities, we have an opportunity to educate within schools and other community-based areas around what socially engineered scams are. Your member is your line of defense,” Postma said.
“Educating and being transparent about what’s going on and playing a critical role, even outside of interactions that happen at the financial institution,” she said. “That becomes our most effective method.”
