What you missed at EuroFinance 2026: our roundtable on automation and AI in FX risk management
EuroFinance’s 35th International Treasury Management conference brought more than 2,600 treasury professionals to Barcelona from 16 to 18 September. AI was the headline topic across the three days, and it drew a full room to the AtlasFX roundtable, Managing corporate FX risk with automation and AI, where treasurers from multinational corporates across a range of industries compared notes on how they manage currency risk today.
If you couldn’t be there, here’s what you missed. The conversation was candid, practical and, in some ways, surprising: for all the attention on AI, the discussion kept coming back to the fundamentals of how FX risk is managed day-to-day.
1. Automation maturity varies widely, and the gaps start early.
Every FX hedging decision depends on something that happens long before a trade is placed: gathering accurate exposure data from across the business. For multinational companies, that data typically lives in multiple ERPs, subsidiaries and systems, and has to be collected, validated and aggregated before treasury can act on it.
Around the table, it was clear that companies are at very different stages of automating this process. Some have made real progress, while many teams still rely on manual steps to collect and check their exposures. That manual effort comes at a cost. It slows the process down, consumes time that could be spent managing risk, and can introduce control gaps before a hedging decision is even made.
Part of the challenge is that ERP systems are rarely configured with treasury’s needs in mind, which leaves many teams piecing exposures together by hand. When teams fill those gaps with spreadsheets, risks can build quietly and often go unnoticed until something goes wrong.
2. Explaining FX results is still harder than it should be.
Once hedges are in place, treasury has to explain the outcome to the CFO, the board and other stakeholders. That’s where many teams struggle. A net FX gain or loss is the product of several moving parts, and separating them is not straightforward.
The group identified a clear need for better analysis that can attribute results across forward points, spot-rate movements, execution slippage, unhedged or under-hedged exposures and accounting mismatches. Without that breakdown, it’s difficult to tell whether a result reflects the hedging strategy, the quality of the exposure data, market conditions or accounting treatment.
Automation plays a significant role here. When exposures, hedges, market movements and accounting outcomes are connected in one consistent framework, results become far easier to trace and explain, and treasury can close the quarter with confidence rather than scrambling to account for surprises.
3. AI adoption in FX remains cautious.
Given how much of the conference focused on AI, you might expect treasury teams to be racing ahead. The roundtable painted a more measured picture. Participants were tentative about using AI in day-to-day FX risk management today.
The most established use case discussed was machine learning to forecast future exposures from historical data. It’s a natural fit for a problem that depends on spotting patterns in large volumes of information, and one where AI can overcome the limits of traditional forecasting methods. More broadly, the group saw AI’s value in supporting analysis, forecasting and decision-making, rather than replacing the judgement of experienced treasury professionals.
4. When it comes to trading, humans stay in the loop.
The clearest consensus of the session came when the discussion turned to execution. There was strong reluctance to let an AI agent create and execute FX trades without human review.
The preferred model is human-in-the-loop: AI can produce recommendations, flag exceptions and prepare proposed actions, while treasury retains approval and control. For a function built on governance and accountability, that balance matters for both internal confidence and auditors.
The bottom line
The roundtable’s conclusion was clear. The immediate opportunity in FX risk management isn’t fully autonomous trading. It’s the intelligent automation of data collection, exposure forecasting, hedge analysis and results attribution, supported by strong controls and appropriate human oversight.
That message echoed through our conversations at the AtlasFX stand throughout the week. AI is part of the conversation, but workflows, automation and data remain at the heart of what treasury teams want to solve right now.
Continue the conversation
If you missed the roundtable, we’d welcome the chance to discuss how your team is approaching these challenges, from exposure collection through to results attribution.
We look forward to seeing you in Amsterdam for EuroFinance 2027!
