CFOs That Already Love Gen AI Are Most Ready for Agentic AI

When the term ‘agentic AI’ first began circulating in corporate circles, most chief financial officers (CFOs) regarded it as a futuristic curiosity, an intriguing but distant idea. Yet from then until now, the conversation has already shifted. And quickly.

According to data in the September 2025 edition of The CAIO Report from PYMNTS Intelligence, “How Agentic AI Went From Zero to CFO Test Runs in 90 Days,” nearly 7% of U.S. enterprise CFOs have already deployed agentic AI in live finance workflows, while another 5% are running pilots. The velocity of this shift has few parallels in the technology adoption history of the finance office.

The change reflects both the hunger among CFOs for tools that can ease rising cost pressures and reporting demands, and the growing sophistication of AI platforms that promise to act as autonomous decision-makers rather than mere analytics engines.

What emerged clearly from the PYMNTS Intelligence data was that companies already successfully using gen AI are creating organizational buy-in for agentic. Businesses that saw the most positive return on investment (ROI) from gen AI are also proving the quickest to adopt its agentic offshoot.

But the story is not simply one of unbridled enthusiasm. The PYMNTS Intelligence report highlighted a two-track response: a front-footed exploration of agentic AI’s potential benefits and a back-footed reluctance to surrender data access or authority over core treasury, compliance, and payment decisions.

That tension will likely define the next phase of adoption.

Finance Culture Meets Agentic and Autonomous Intelligence

The term agentic AI describes systems that can perceive data flows, interpret context, and take actions—ranging from scheduling workflows to executing low-risk transactions—without constant human prompting. These agents often integrate large language models with rules-based process engines and have the capacity to operate across multiple enterprise software environments.

The early deployments have focused on relatively low-risk, insight-driven tasks. Agentic AI is being used to generate rolling forecasts, synthesize large datasets for scenario modeling, and prepare draft versions of management reports. Some firms have also experimented with letting agents recommend cost-containment strategies or alert teams to compliance anomalies in near real time.

Yet the leap from recommendation to action remains rare. The study found that none of the surveyed CFOs were ready to give their agentic tools full, unfettered access to company data and permission to execute transactions. Only 8.3% were willing to allow even moderate levels of autonomy. The vast majority preferred a human-in-the-loop approach, in which agents can prepare analyses, but humans approve any steps involving money movement or regulatory exposure.

Read the report: How Agentic AI Went From Zero to CFO Test Runs in 90 Days

Overcoming the Trust Barrier

Technology alone does not determine adoption speed; organizational culture plays a crucial role. Finance departments have traditionally prized precision, control and linear reporting lines.

Trust has emerged as the central gating factor. CFOs are charged with protecting financial data, ensuring compliance with evolving global standards, and maintaining a company’s fiduciary posture. For a new class of technology — especially one promising to act without continuous direction — to gain a foothold in that environment, it must demonstrate not only competence but also transparency and control.

Per the report, CFOs appear willing to let agents handle tasks where errors can be caught early or have limited downside, such as drafting planning documents or monitoring expense anomalies. But higher-stakes areas like treasury operations, statutory compliance, and payment authorization remain off-limits until robust guardrails and explainability frameworks are in place.

After all, finance has already absorbed decades of automation in the form of enterprise resource planning and robotic process automation; adding a new layer of semi-autonomous agents raises questions about oversight and accountability.

As with earlier generations of automation, the transition is likely to proceed in stages. Functions that are repetitive, rules-driven and analytically intensive will see the earliest durable adoption. Functions that touch capital movement, external reporting or regulatory exposure will lag until governance and trust frameworks mature.