Plaid CEO Zach Perret warned that artificial intelligence’s rapid expansion could create a “scary year” for financial security. Speaking at the Money20/20 conference in Las Vegas on Tuesday, Perret described how large language models like OpenAI’s ChatGPT and Anthropic’s Claude may soon act as financial intermediaries rather than just analysis tools.
Monetization Through Embedded Finance
Perret suggested that AI platforms will likely become “harnesses” for financial services. These platforms would embed card issuing capabilities, collecting interchange fees directly through the AI interface. The Plaid co-founder believes this will be a logical first step, though he also expects more complex lending products to follow.
As these AI tools move from analysis to action, the lines between traditional banking and AI-led services will blur. Perret noted that the industry is currently debating whether customers will keep their entire financial lives inside these AI harnesses or if banks and fintechs will build their own “agentic experiences” to compete for that space.
Some financial institutions are already preparing for this shift. Fifth Third CEO Tim Spence said his bank plans to integrate services into its apps that consumers can access through Meta’s personal AI agent, Muse. Spence emphasized the importance of offering “bank-grade information security” in these new environments.
Fraud Risks and Industry Response
Fitch Ratings recently assessed the impact of AI-related business disruption on banks over the next five years. The agency classified the risk as moderate, noting that most exposure falls on institutions serving households and small and midsize businesses. These banks typically offer standardized, high-volume products that are easier for AI to replicate.
Read Also: JPMorgan, Goldman Sachs Prioritize Security in AI Expansion
The speed at which customers adopt new service models will determine the extent of this disruption. Perret, however, is less optimistic about the timeline. He suggested that banks will likely respond by building their own agents to maintain control over the customer relationship. A neobank might, for instance, want to handle mortgage conversations within its own “four walls” rather than sending users to an external AI platform.
Fintech chatbots from around 2020 largely failed to gain traction before disappearing, but Perret predicted a new wave is coming. He argued that it is unlikely any single entity will own the entire customer relationship, as people will likely want to handle routine tasks like rebalancing accounts through AI while reserving direct interactions for more complex services like mortgages.
Despite these developments, Perret expressed significant concern over the security implications. He described a Deloitte projection that generative AI could push U.S. fraud losses to $40 billion next year as a conservative estimate. The CEO added that this figure likely excludes the threat of “agent swarms” attacking banks, a scenario he finds particularly terrifying.
Fighting a Technical Arms Race
Perret acknowledged that the industry is engaged in a “dead sprint” to counter these threats. Plaid’s leadership team frequently discusses the frontiers of security and fraud, and the CEO admitted he is “quite scared” by the outlook. He emphasized that the company is actively working to provide customers with security tools and recommendations.
