OvationCXM, the orchestration operating system for financial services, has released new research, AI Agents: From Experiment to Execution, revealing that finance leaders have moved from experimenting with AI agents to using them in their finance workflows — and that they plan to incorporate the technology further into banking and commerce work over the next 12 months.

The survey of 520 U.S. finance leaders found that 75% actively use AI agents, or are piloting or testing them, in at least one finance or banking workflow today. The research covered both current usage and plans for the 12 months ahead. As adoption expands, finance leaders expect their AI agents to move beyond internal tasks and information retrieval to initiating instructions and transactions directly with their financial institutions.

Key findings from the research include:

• 75% are already using or testing AI agents. Finance teams report active use, pilots, or testing across finance and banking workflows.
• Customer AI agents are already touching banking. Three of the top six finance workflows using AI agents involve banking, including transaction tracking, payment approvals, and payment or transfer initiation.
• 68% expect their AI agents to transact. In the next 12 months, finance leaders plan to use AI agents for read-only, instruct, and transact tasks — three escalating levels of agent autonomy that run from retrieving information to executing payments — with 28% expecting to use AI agents for ACH payments and 23% for wire transfers.
• 54% would find a workaround if their bank can't support their AI agents. Among finance leaders already using AI agents, 54% would find an internal or third-party workaround, and 15% would move their banking business elsewhere.

“Customer AI agents are rapidly becoming the next banking channel,” said Alfred “Chip” Kahn, founder and CEO of OvationCXM. “The opportunity for banks is to lead this shift by becoming the trusted gateway for how those agents interact with their systems. That means knowing whose agent is making the request, what it is allowed to do and when a human needs to be in the loop.”

The research also shows that finance leaders are not asking for unrestricted AI autonomy. Their top confidence-building requirements center on permissions and limits, the ability to suspend or revoke agent access, real-time visibility, audit and reconciliation history, and identity verification. When human approval is part of the process, finance leaders are nearly as willing to let their agents initiate or advance trans-level banking tasks as they are read-only activities — 50% compared with 54%. Those requirements track closely with the gateway role Kahn describes, in which a bank knows which agent is making a request, what it is permitted to do, and when a human has to be involved.

“If a bank can't support its customers' AI agents, those customers will find another way to get the work done, either with their bank or somewhere else,” Kahn said. “The risk is that more of the customer experience moves to a competitor or outside the bank. Banks need to establish this control point now as the channel takes shape.”

With 68% of respondents expecting transact-level agent use within the next 12 months, the survey positions bank readiness for customer agents as a factor in where finance teams maintain their banking relationships — a dynamic the workaround and switching findings put in concrete terms.

Source: Global FinTech Series