Agentic AI in Treasury: How Leaders Balance Speed, Control, and Confidence

Artificial Intelligence (AI) is rapidly evolving from an analytical support tool into something more autonomous with the introduction of Agentic AI, which initiates actions, adapts to conditions, and recommends next steps. For treasury teams managing liquidity, payments, and financial risk, this evolution promises efficiency.

However, unlike traditional automation, agentic AI introduces new governance challenges. When systems begin to act on behalf of finance teams, data quality, accountability, and transparency become mission-critical.

In a recent SAPinsider Expert Exchange, Dory Malouf, Senior Director of Global Value Engineering and Felix Grevy, Senior Vice President of Platform, Data, and AI at Kyribatold Susan Galbraith, Vice President and Research Analyst at SAPinsider how treasury teams are preparing for AI systems that can act and not just advise..

Autonomy Requires Boundaries

Malouf cautioned against equating autonomy with value. “Treasury teams are under constant pressure to move faster,” he said, “but speed without control doesn’t create confidence—it creates risk.” From Kyriba’s perspective, AI must increase transparency and oversight, particularly when decisions affect liquidity and counterparty trust.

Grevy expanded on the technical foundations required for responsible autonomy. “Agentic AI needs a complete understanding of the financial environment,” he explained. “That means unified data across cash, payments, bank connectivity, and historical behavior.” He also warned that fragmented systems undermine AI’s ability to operate safely.

Designing for Humans, Not Just Systems

Governance emerged as a central theme throughout the discussion. As AI systems gain autonomy, organizations must define what AI can do independently, when human review is required, and how decisions are documented. Grevy emphasized that governance is not a brake on innovation. “Trust allows you to scale AI with confidence,” he said.

The Kyriba experts also addressed usability. Treasury users increasingly expect intuitive, consumer-grade experiences, even while managing complex enterprise processes. Agentic AI can simplify interactions but only when it operates within clearly defined controls. Malouf stressed that human judgment must remain central, particularly during market volatility or operational disruption.

Looking ahead, Kyriba’s leaders framed agentic AI as an evolution rather than a flip of the switch. Organizations that invest early in integration, governance, and data quality will be best positioned to adopt autonomous capabilities responsibly.

What This Means for SAPinsiders

Agentic AI forces treasury leaders to formalize decision authority. The shift from AI-generated insight to AI-initiated action fundamentally changes accountability in treasury. When systems can trigger workflows, recommend liquidity moves, or flag exceptions autonomously, organizations must clearly define which decisions AI is permitted to execute and where human approval is mandatory. Thus, SAP finance leaders should assess whether their existing control frameworks, approval hierarchies, and audit trails are designed for this level of autonomy. Without explicit decision boundaries, agentic AI can introduce governance gaps that increase audit and regulatory risk rather than reduce operational friction.

Integration is no longer optional in an agentic world. The discussion made clear that autonomous AI cannot function responsibly on partial information. Agentic systems require a unified, real-time understanding of cash positions, bank connectivity, payment flows, and historical behavior to act with confidence. SAPinsiders who operate in fragmented landscapes where treasury data is split across ERP, bank portals, and point solutions should prioritize integration over AI expansion. Otherwise, organizations risk deploying AI agents that base decisions on incomplete or outdated context, which undermines trust and limits the practical value of autonomy.

Human oversight becomes more valuable—not less with AI. Agentic AI does not diminish the role of treasury professionals. It increases the importance of human judgment. As Malouf and Grevy emphasized, AI is most effective when it handles execution and pattern recognition while humans manage interpretation, exceptions, and accountability. SAPinsiders should design AI initiatives that deliberately keep humans in the loop, particularly during volatility or disruption. The most successful treasury teams will be those that position AI as an augmentation layer—freeing experts to focus on risk assessment and strategic guidance while remaining accountable for outcomes.