Key Takeaways
- Banks have invested heavily in data‑infrastructure (APIs, consent frameworks, secure sharing) but have largely left the data idle, treating the build as a compliance exercise.
- Agentic AI can now act on that infrastructure, performing multi‑step reasoning and executing transactions without human initiation for each step.
- Fiserv’s newly launched agentOS provides a single, governed operating system for deploying and managing AI agents across core banking, payments and servicing workflows; six banks helped design it, two are already in beta, and OpenAI and AWS are collaborators.
- Early pilots show concrete efficiency gains: First Interstate Bank’s commercial‑loan‑onboarding agent replaces a multi‑system, manual process, while Boulder Dam Credit Union’s operational‑analysis agent cuts report generation from ten minutes to seconds.
- The platform includes kill‑switches, human‑in‑the‑loop controls and audit trails to satisfy bank‑grade regulatory requirements.
- The Financial Data Exchange (FDX) is developing standards for AI‑agent access to consumer financial data, focusing on authorization, data scope, permission tracking and liability—issues not covered by existing consumer‑data‑sharing rules.
- FDX CEO Kevin Feltes stresses that broad industry collaboration is essential to protect consumers while enabling innovation; the call for input runs through May 29.
- The central unresolved question remains who captures the value created when AI agents activate dormant data pipelines—banks that built the infrastructure, fintechs that have long used the data, or platform providers like Fiserv.
Overview of the AI Infrastructure Challenge
Banks have spent years constructing the “data pipes” that move information securely across systems—APIs, consent frameworks, and robust sharing protocols. Yet, as the article notes, many institutions treated this build as a compliance checkbox: “They built the APIs and stopped. Data flows, but very little happens with it.” The emergence of agentic AI now adds an active layer capable of reasoning across those pipes, initiating actions, and completing transactions without waiting for a human to trigger each step. This shift transforms dormant infrastructure into a potential engine for efficiency and new revenue, but it also raises fresh questions about governance, liability, and value capture.
Fiserv’s AgentOS: A Unified Operating System for AI Agents
Recognizing the gap between idle data and AI‑driven action, Fiserv introduced agentOS, described in the piece as “an operating system designed to let financial institutions deploy and manage AI agents across core banking, payments and servicing workflows from a single governed environment.” Six banks co‑created the platform, with two currently in beta, and technology partners OpenAI and AWS signed on as collaborators. Dhivya Suryadevara, Fiserv Co‑president, explained that every bank client the company has spoken with faces “cost, deposit competition and a retiring workforce,” and agentOS is positioned as a simultaneous answer to those pressures. The platform’s success will be tested over the next twelve months as it attempts to scale across thousands of institutions and the heterogeneous systems Fiserv does not directly control.
Real‑World Pilots Demonstrating Tangible Benefits
Early use cases illustrate how agentOS can translate theory into practice. First Interstate Bank is piloting an agent for commercial loan onboarding—a process that “currently spans multiple systems and requires significant manual hours.” By delegating data gathering, validation, and preliminary underwriting to an AI agent, the bank hopes to slash processing time and free staff for higher‑value interactions. Meanwhile, Boulder Dam Credit Union runs a daily operational analysis agent that “compressed report generation from ten minutes to seconds.” These examples underscore the efficiency gains possible when AI agents can autonomously traverse the data infrastructure banks have painstakingly built.
Governance Controls Built into AgentOS
To satisfy the stringent regulatory expectations of banks, Fiserv has embedded several safeguards into agentOS. The article highlights that the platform includes “kill switches, human‑in‑the‑loop controls and audit trails designed to meet bank‑grade regulatory requirements.” Kill switches allow immediate cessation of an agent’s activity if anomalous behavior is detected; human‑in‑the‑loop controls ensure that critical decisions—such as loan approvals or large fund transfers—receive final human oversight; and immutable audit trails provide regulators with a verifiable record of every action taken by an AI agent. These features aim to balance innovation with the risk‑averse culture prevalent in financial services.
The Regulatory Vacuum: Who Authorizes AI Agents?
As AI agents begin to act on behalf of consumers, existing consent mechanisms fall short. The Financial Data Exchange (FDX) has launched an initiative to address the structural gap: “When a consumer connects a bank account to a third‑party app, the consent is visible and deliberate. When an AI agent does the same thing on a consumer’s behalf, the questions multiply: who authorized the agent, what data can it access, how is that permission tracked and who is liable when something goes wrong.” The current standards governing consumer financial data sharing were never drafted for autonomous agents, leaving a regulatory gray area that could expose both firms and consumers to unintended risks.
FDX’s Call for Industry Collaboration
FDX CEO Kevin Feltes emphasized that solving these challenges will require “broad industry collaboration” to ensure connections are built in a way that protects consumers. The organization opened a call for input running through May 29, inviting banks, fintechs, technology providers, and consumer advocates to shape the forthcoming guidelines. The outcomes will determine what banks and fintechs can actually deploy and the degree of regulatory exposure they assume when they do. In essence, FDX is attempting to draft the “rules of the road” before AI agents start traffic on the data highways banks have paved.
The Value‑Capture Dilemma
Beyond safety and compliance lies a more fundamental economic question: who reaps the benefits when AI agents unlock the value stored in years‑old data infrastructure? The Camunda analysis, drawing on Datos Insights research, observes that many banks treated their API and consent investments as a compliance exercise, leaving the data largely untapped. Agentic AI changes that calculus by enabling “an agent [to] reason across multiple systems, sequence a set of actions and execute a transaction without waiting for a human to initiate each step.” Consequently, both the banks that built the pipes and the fintechs that have long leveraged the data stand to gain, as do platform providers like Fiserv offering the orchestration layer. The article concludes that “Neither question has a settled answer yet,” highlighting that the next year will be critical in clarifying roles, responsibilities, and profit sharing in the AI‑enabled banking ecosystem.