Regulatory Roadblocks: Why Agentic AI Faces Unique Challenges in Banking
- Agenic AI has impressed many in the banking industry
- Meticulous care is required when handling customer data
- Today's AI is not agenic AI
As financial institutions gravitate toward innovative automations like agentic AI, the allure of “zero operations”—where everyday banking processes run themselves—appears within reach. Yet, according to Biswajit Das in his recent Forbes Technology Council post, regulatory and compliance hurdles remain the most formidable barriers, casting serious doubt on the suitability of agentic AI for banking’s uniquely high-stakes environment.
Heavily Regulated Terrain
Banking is one of the world’s most tightly regulated industries, Mr. Das notes. Any technology that handles sensitive customer data and navigates compliance requirements must do so with meticulous care. Agentic AI, which relies on massive datasets and adaptive models, directly amplifies long-standing concerns about privacy, cybersecurity, and data accuracy. The specter of AI “hallucinations”—erroneous or unpredictable outcomes—combined with a lack of clear regulatory frameworks, makes it essential for banks to "bake in" compliance, auditability, and governance from the ground up.
Traditional automation will remain powerful and relevant, but AI agents bring a new dimension: They can leverage generative AI to drive workflows, combining rules-based automation (UI- or API-based) with AI-enabled decision making. Before rushing into the “agentic AI” revolution, banks must take a deliberate, human-centric approach.
Major players like JPMorgan Chase have responded by enacting strict guidelines for reviewing AI models prior to deploying them. New watchwords like “explainable AI,” “responsible AI,” and “ethical AI” have shifted from optional upgrades to nonnegotiable standards, particularly in contexts where violations carry severe legal and reputational risks.
Governance and Human Oversight
Forbes highlights the emergence of “human-in-the-loop” mechanisms—checkpoints where trained staff verify AI-generated decisions before they impact customers. These are not mere technical features but compliance necessities: without human oversight, agentic AI could make credit or fraud decisions out of step with regulation and ethics.
While agentic AI promises radical operational efficiencies, Mr. Das argues that banks must approach this revolution with caution. Embedding regulatory controls is not a mere add-on but a foundational requirement. Until the industry and regulators can ensure that agentic AI operates safely, reliably, and transparently at scale, its adoption will be limited—not by technology, but by regulatory prudence.
AI for Check Processing and Fraud Detection
As you are aware, financial institutions are currently deploying AI for check processing and fraud detection. To be clear, the AI utilized today is NOT agentic AI. The real question is: Do FIs need agentic AI for these processes?
Unfortunately, the answer is not a simple yes or no.
When we assess check processing automation, FIs are achieving over 99% accuracy and read rates. These results have automated the processing of checks and, without needing a system for decisioning, have achieved straight-through-processing. So there isn't a true need for agentic AI.
On the check fraud detection side, FIs need to proceed with caution. Agentic AI is not at the point for automating decisioning on fraudulent items. The current AI technologies are achieving 95% detection, but is only achieved through a multi-layered technology approach where checks are analyzed by multiple technologies that serve a specific purpose such as behavioral analysis of the account, transactional analysis, and image forensics to analyze the images of checks for inclearings and deposit. These systems identify issues with the check transactions and provide risk scores that are reviewed by a trained fraud analyst.
Can agentic AI replace the manual review? That is a question that banks will need to ask themselves and assess the risk. However, in today's environment, it is not recommended.
Agentic AI will have its place in financial institutions. Though, with the regulatory hurdles, we are years -- possibly decades -- away from this becoming a reality.