Agentic AI: Banks Face QA Nightmare Over Unexplained Systems

Banks are facing a growing problem with agentic AI: many technology vendors are selling these systems without adequately explaining how they work. This is becoming a major governance and regulatory testing issue for banks' QA and software testing teams. Ron Shevlin, Chief Research Officer at Cornerstone Advisors, notes that vendors focus on benefits like speed and efficiency. He asked Claude to review vendor websites and found they emphasize outcomes like faster decisions and reduced analyst time, rather than explaining the underlying processes. The challenge for banks is understanding how these AI systems reach their outcomes. This means QA and testing teams must now validate how AI makes decisions, handles exceptions, and ensures human involvement and governance controls. Regulators are taking an aggressive stance on AI governance, requiring banks to explain model governance and credit decisions. Vendor claims about "80% faster decisions" don't help banks answer these critical questions from boards, regulators, and compliance teams. This situation leaves banks exposed when operational questions arise.