The most powerful technology for committing financial fraud today isn’t in the hands of security teams—it’s being wielded by the fraudsters themselves, in a chilling cat-and-mouse game where the “mouse” is currently winning. This startling reality frames a wide-ranging discussion on the rollercoaster journey of fintech, tracing its path from a niche “startup industry” to a force now synonymous with financial services itself. The conversation, featuring Plaid’s Zach Perret and a16z’s David Haber, maps the sector’s dramatic seasons: the “late spring” of 2018-2019, the “EDM-pumping summer” of hyper-growth during COVID, the deep freeze of “fintech winter” in 2022-23, and the cautious, green-shoot recovery of today.

This cyclical journey has fundamentally reshaped the landscape. The initial wave was about solving access—digitizing core banking products like checking accounts, loans, and investments so they were available outside a physical branch. That mission is largely accomplished. Now, the focus has shifted to making these digital services excellent and ubiquitous. Excellence means tackling endemic problems like illogical credit scoring and, most pressingly, the AI-fueled fraud epidemic. Ubiquity means “embedded finance,” where financial tools appear seamlessly in non-financial contexts, from “Buy Now, Pay Later” at any retailer to banking services embedded in companies like John Deere or Ford.

Looking ahead, the industry is converging around a few key themes. AI’s dual-edged nature is paramount: while it powers sophisticated scams like AI-driven “pig butchering,” it also offers hope for better underwriting, automated customer service agents, and finally realizing the long-promised vision of “self-driving money.” Furthermore, the posture of large incumbent banks has shifted from building everything in-house to becoming eager buyers of external software, especially AI tools that can automate manual back-office work. The survivors of the boom-and-bust cycle—companies like Robinhood, SoFi, and Revolut—have emerged as full-stack, rebundled financial giants, stronger and more diversified than before.

Surprising Insights

• AI’s biggest financial services use case is currently fraud: The most effective application of AI in finance right now isn’t customer service or investment advice, but fraudsters using it to scale scams like “pig butchering” (building trust via AI chatbots to steal money).
• “Fintech winter” created stronger, long-term companies: The funding drought forced a shakeout, eliminating point-solution neobanks and unsustainable lenders. The survivors were pressured to expand into full-stack offerings (adding lending, investing, etc.), resulting in more durable, diversified businesses.
• Major banks have radically changed their tech philosophy: Institutions like Goldman Sachs, which once famously built their own email client, have undergone a “humbling” and are now actively seeking to adopt best-in-class external software, especially AI, to solve workflow problems.
• The future of credit scoring may ignore traditional credit reports: New models like Plaid’s “Lens” score focus on real-time income and expense data to assess loan risk, arguing that a recent pay raise or responsible spending is more logical and indicative than a historical repayment file.

Practical Takeaways

• For entrepreneurs: Consider building B2B software that automates manual processes (compliance, treasury, loan servicing) for large financial institutions. Their appetite for AI-driven efficiency is higher than ever, and sales cycles are accelerating.
• For consumers: Be profoundly skeptical of unsolicited digital communication. The “pig butchering” scam starts with a simple “Hey, how you doing?” text; engaging can lead to an AI-driven campaign designed to steal your money through engineered trust.
• For product builders: If creating AI “agentic” financial tools, prioritize extreme transparency and user education. A power user might trust an AI to sweep money between accounts, but mainstream users will need clear explanations for every action to build comfort.
• For the industry: Invest heavily in cross-institutional, network-level fraud detection. Fighting AI-powered fraud requires aggregated data across many companies to identify anomalous patterns, moving beyond siloed, company-by-company defenses.