Bob Stark, global head of Market Strategy at Kyriba—winner of two global FX Tech awards—says lack of data visibility, not a lack of AI tools, is costing companies billions in currency losses.
Global Finance: Kyriba’s data often highlight billions in corporate foreign exchange losses. What is the most prevalent mistake companies make?
Bob Stark: The biggest thing we find is that they don’t have enough data to make a data-driven decision around FX, which boils down to a lack of visibility into their cash flow and balance sheet exposures.
Insufficient data and inadequate technology prevent accurate understanding of true exposures, making informed hedging difficult for both balance-sheet and cash-flow exposures. The result is uncertainty in hedging, often leading to underhedging or overhedging, which can significantly impact an organization’s P&L.
GF: To what extent are corporate treasurers effectively leveraging AI and advanced analytical tools in managing FX risk data?
Stark: AI tools like Co-pilot and ChatGPT are increasingly used by corporate finance teams, primarily for outside sentiment analysis—analyst or bank commentary, rate curves, etc.—to gain an external view. According to a recent study from Gartner, 59% of corporate finance teams use some flavor of AI, with knowledge management being the most common use case. This suggests to me that most corporate finance teams are not fully operationalizing AI.
The key challenge is still data. While external sentiment is helpful, the most valuable data is internal. Effective FX-risk management requires a robust data strategy first, ensuring complete, reliable data to quantify currency impact on balance sheets, cash flows, and earnings. This strategy must support both operational and communication needs.
Actioning this knowledge involves either using derivatives or pursuing natural hedging. Even advanced analytics solutions like Kyriba’s predictive analytics will lack effectiveness without this foundational data strategy. You can’t have an AI strategy until you have your data strategy figured out.
GF: Kyriba’s trusted agentic AI (TAI) uses a human-in-the-loop system to flag FX anomalies or forecast variances, among other capabilities. For your 2026 self-service agent builder road map, will it have capacity for autonomy?
Stark: AI can handle the entire FX journey, from exposure identification and quantification through to hedge recommendation, execution, and back-office compliance, including valuations and accounting entries. However, full end-to-end AI is not a present reality for any treasury operation because a human should remain in the loop.
Organizational adoption of AI depends on trust, which is earned over time through experience, transparency, and proven data security. Organizations must find the right balance between AI and human tasks, as this balance varies by organization and AI maturity.
I advise organizations to visualize their workflow—e.g., using Kyriba’s menu maps—and assign specific tasks to either AI agents or humans. This approach breaks end-to-end processes, like FX hedging, into manageable, realistic, sequential pieces. This gradual, piece-by-piece approach is about building comfort and trust.
We named our initiative Trusted AI because trust is the key differentiator. It is unrealistic to expect immediate, full-scale adoption without this earned trust.
GF: How does Kyriba help clients transition from basic data management to a sophisticated, AI-ready data strategy?
Stark: Our role is not necessarily to tell customers what to do, but to ensure they have the data to support their decisions. For US dollar-based organizations that historically didn’t need to focus on this level of currency exposure, corporates are increasingly realizing the opportunity to improve data strategies. For all our customers, whether in Europe, the Americas, Central and Eastern Europe, the Middle East, and Africa, or Asia-Pacific, having a more intentional data strategy is a meaningful step to improve the maturity of their hedging and currency operations. This robust data foundation is the essential precursor to successfully implementing a more AI-driven hedging program.
