Why Finance Leaders Trust Generative AI but Hesitate on Autonomous Agents

Transparency, Not Speed, Could Decide AI’s Future in Finance

Corporate finance has consistently been an early adopter of automation. From the days of Lotus 1-2-3 to robotic process automation (RPA), finance professionals have welcomed tools that reduce manual effort while ensuring strict governance. Generative artificial intelligence (AI) fits neatly into this tradition.

According to findings from the July 2025 PYMNTS Intelligence Data Book, “The Two Faces of AI: Gen AI’s Triumph Meets Agentic AI’s Caution,” CFOs are enthusiastic about generative AI. Nearly 9 in 10 report strong ROI from pilot projects, and 98% feel comfortable using it for strategic planning.

However, the mood shifts when discussions move from AI as a copilot or dashboard assistant to fully autonomous “agentic AI” systems—software that acts independently, makes decisions, and executes workflows without human supervision. Only 15% of finance leaders are considering its deployment. This hesitation reveals a deeper conflict: a legacy system built to minimize risk versus new AI systems built to take action.

Why Agentic AI Feels Different

Generative AI won over finance leaders because it makes work easier without breaking existing rules. It speeds up analysis, drafts explanations, and highlights risks—all while humans retain the final say. This aligns perfectly with the finance function’s long-standing priorities: faster closes, improved forecasts, and doing more with less.

Agentic AI, in contrast, doesn’t just suggest; it acts. It can reconcile accounts, process transactions, or file compliance reports automatically. This autonomy raises serious concerns. Executives comfortable with AI writing reports become cautious when AI starts moving money or approving deals.

• Governance: Who is accountable when a machine transfers funds?
• Visibility: Security teams may lack insight into what autonomous AI is doing once it accesses systems.
• Accountability: Regulators won’t accept “the software decided” as a valid explanation for errors in tax filings.

The black-box nature of AI complicates matters further. Unlike traditional rule-based systems, agentic AI relies on probabilistic reasoning, often without clear audit trails. For finance leaders who must explain every figure, this lack of transparency is a deal breaker.

Legacy infrastructure adds to the challenge. Finance data is scattered across multiple platforms—enterprise software, procurement tools, banking portals. For AI to act autonomously, it needs seamless access across these systems, navigating complex authentication and siloed permissions. Managing this for machines acting like employees, but faster and harder to monitor, is a significant hurdle.

The Path Forward: Transparency as the Priority

For autonomous AI to move beyond pilot projects, it must deliver measurable value. CFOs want to see reduced cycle times, fewer errors, and improvements in working capital. They expect audits to be smoother, not more complicated.

The key isn’t perfection, but explainability. Transparency is the critical feature agentic AI must provide to gain trust. Without it, these systems risk remaining ideas rather than becoming integral parts of finance operations.

Finance professionals interested in expanding their knowledge on AI applications can explore practical courses and certifications at Complete AI Training.