A recent survey found that only 43% of CFOs trust their organisation’s AI governance. The next wave of finance automation involves agentic AI, systems that can execute multi-step processes, make decisions within parameters, and interact with other systems without waiting for human approval at each stage. Governance now has to address whether the rules around an autonomous process can catch errors before they compound.
We have started hearing this concern surface in senior finance searches across the DACH region. Audit committees want to know who owns the controls when an AI agent processes intercompany reconciliations or flags exceptions in treasury workflows. The CFO is the natural answer, but many finance leaders we speak with admit they lack the technical fluency to evaluate what an agent can and cannot do. They can read a dashboard, but they cannot interrogate the logic underneath it.
Companies are responding by looking for finance professionals who can sit between the data science function and the control environment, people who understand both the accounting risk and the system architecture. A controller who can articulate what happens when an agent encounters an edge case in transfer pricing carries more weight in hiring conversations than one who simply trusts the vendor’s assurance. A treasurer who can explain how an AI-driven cash positioning tool interacts with bank APIs stands out from one who approves the rollout without asking.
The 43% figure is a snapshot of a moment when most AI in finance still requires human checkpoints. Agentic systems remove some of those checkpoints by design. Finance leaders preparing their teams need someone on staff who can explain how the guardrails work and where they might fail.
Prompted by reporting from CFO Dive.