The state of FX risk management: what treasury pros are actually saying
Spend any time at treasury conferences or in rooms with FX risk managers, and a pattern emerges quickly. The keynote narrative is all about transformation: AI-powered forecasting, real-time visibility and seamless integration. The hallway conversation is more honest. Most teams are still fighting the same battles they were two or three years ago, just with more pressure to fix them faster.
We’ve been in a lot of those rooms over the past year: private roundtables, conference floors and hallway conversations at ACT, EuroFinance, SFTS and Nordic 360. The same themes keep surfacing. Here’s what we’re hearing.
Most FX teams are still running on manual processes, and they know it
Ask a room full of FX risk managers how they’re approaching exposure forecasting, and the dominant answer isn’t AI or sophisticated statistical modeling. It’s Excel and judgment calls. Most treasury teams are working through either mostly manual processes or system-generated outputs that require significant manual adjustment before anyone will act on them.
That’s not a criticism. It’s reality. FX exposure forecasting involves a lot of nuance: business unit relationships, deal timing, policy interpretation. It doesn’t lend itself neatly to automation, which is partly why progress feels slow even when the appetite for improvement is high.
What’s notable is that very few teams have nothing. Most have built a process, even if that process is a spreadsheet someone maintains carefully. They’re not starting from zero. They’re trying to figure out how to evolve what they have without breaking it. The bottleneck, as we explore in our FX Risk & Treasury Challenges whitepaper, isn’t always tools. It’s often the process maturity and data quality those tools depend on.
AI pressure is real, but so is skepticism
Whether it’s coming from the CFO, the board or the general noise in the industry, almost every treasury team is fielding some version of the same question right now: what are we doing with AI? The pressure to have an answer is real. The path to a good one is murkier.
Ask those teams what’s actually standing in the way, and you get a long list: integration challenges, governance concerns, resource constraints, data quality issues and a basic distrust of AI outputs. No single thing dominates. That’s actually the important detail. If there were one obvious blocker, you’d fix it and move on. When everything is a barrier, the problem isn’t picking the right tool. It’s whether the conditions exist for the tool to work.
Of all those barriers, distrust of model outputs consistently ranks highest when teams are asked to name their single biggest obstacle. That’s worth sitting with. AI trust isn’t a technology problem. It’s an organizational one, and no new model release fixes it.
Data quality is the unglamorous prerequisite
Data quality is the other one nobody wants to own. It’s the problem everyone knows about and nobody wants to talk about at a conference because it’s not exciting. But bad, incomplete or inconsistently structured data is why so many AI and automation initiatives never make it past the pilot stage.
For FX specifically, exposure data tends to sit across multiple systems at once: ERPs, TMS platforms, spreadsheets and email threads, with nothing connecting them cleanly. Before AI can help you forecast or hedge more effectively, you need to trust what you’re feeding it. That’s a data problem before it’s a technology problem, and it’s one of the most consistent issues we see in FX risk management programs.
The biggest opportunity? Sharper FX exposure data, smarter hedging decisions
Ask treasury professionals where their function could make the most difference to the business, and the answer keeps coming back to the same place: FX exposure visibility and the hedging decisions that depend on it. A team that knows what it’s actually exposed to, across currencies and entities, and builds hedging strategies on that foundation stops reacting to bad news. It starts preventing it.
Cash flow and balance sheet hedging programs built on best-guess exposure estimates are different from ones built on clean, integrated data. One holds up when leadership asks hard questions. The other creates more of them. Get the exposure data right and the downstream work gets materially easier: hedge design, effectiveness measurement, reporting to leadership.
It’s also where AI is starting to earn its keep in FX. Tools that use machine learning to model historical exposure patterns, flag anomalies and sharpen forecasts are past the pilot stage in some organizations. Still one piece of a bigger workflow, but a real one. But like all models, there is the potential for GIGO, or “Garbage In Garbage Out” that AI can’t overcome. Coming to a flawed conclusion faster doesn’t help anyone. Expertise is still crucial when it comes to understanding what issues reside in FX exposure data within the ERP, and how to fix them.
If you want to go deeper on where AI fits in a mature FX program, our definitive guide to FX risk management is a good place to start
Manual work is the strategic ceiling
Across all these conversations, one constraint surfaces more than any other. Nearly half of treasury respondents identify manual work as their primary barrier, and the consequences follow fairly directly.
If your team spends most of its time pulling data, reconciling reports and building decks, there’s not much left for the work that justifies having a treasury team in the first place: stress-testing hedging strategies, engaging business partners on FX exposure or advising on capital allocation decisions.
Most treasury teams have started modernizing. Very few are done. The majority describe themselves as somewhere in the middle: some automation, some visibility, but not yet fully connected or running in real time. That’s where most organizations actually are right now.
The right order of operations
The pressure to move on AI isn’t going away, and most teams know it. What’s less obvious is the right sequence: which foundational problems to fix before AI becomes part of the equation.
The teams making the most progress aren’t the ones with the most budget or the most aggressive timelines. They’re the ones that sorted out the basics first. Clean exposure data. Systems that actually connect. Governance that makes outputs trustworthy rather than just fast. That’s the foundation AI actually runs on.
For FX teams specifically, the near-term priorities are clearer than the broader AI conversation might suggest: sharper exposure forecasting, cleaner data, better hedge effectiveness measurement. None of that requires a transformation project. All of it is what makes AI useful once you’re ready to use it.
Nobody in these conversations has this fully figured out. But the teams making real progress share a pattern: they started with the unglamorous work. Clean data, connected systems, governance that makes outputs trustworthy. That’s the foundation. And there’s still room to build it.
