At the recent EuroFinance International Treasury Management conference in Budapest, corporate treasurers exchanged experiences on cash forecasting, FX operations, payments, treasury automation and data architecture. Artificial intelligence was presented there as a tool to generate actionable information for management teams more quickly while reducing manual interventions. Cash forecasting and process automation were the most highlighted applications of AI by participants, but other areas were also presented, notably FX and the coding capabilities that allow information systems to talk to each other.
A pulse check of European treasury teams
Asked in session which activities AI would be useful for, treasurers cited cash forecasting first (38% of responses), followed by process automation (32%). However, more than half of respondents said their company still does not use AI for forecasting, while nearly a third said they are “exploring” how to use this innovation.
92% of participants in the session dedicated to “AI & Automation” indicated that they use AI in their daytoday treasury activities. This very high percentage may reflect a sophisticated—if not selfselecting—panel attending the session. But when the question turned to how AI was being used, the answers were more downtoearth: drafting and polishing emails, summarizing long threads and presentations, and analyzing data that would have taken hours to prepare just a year ago.
AI in treasury processes
The various sessions provided an overview of how treasury teams are experimenting with AI. Kathy Brustad, Director, Global Treasury and Financial Services at Microsoft, explained how her treasury department uses machinelearning models to help predict late payments and strengthen its cash forecasts, while using generative AI to convert plainEnglish questions into SQL to query structured datasets. The result: shorter forecasting cycles, fewer routine errors and more time devoted to decisionmaking.
Microsoft is not the only treasury to have experimented with AI to improve forecasting reliability. A year ago, Redbridge wrote about the longterm project pursued on the same topic by Groupement Les Mousquetaires. In both cases, teams deliver the same lesson: to reinforce cashforecasting processes with AI, you need clean data, clear governance and a sharpeyed specialist to correct model and dataset bias and drift in near real time.
FX and hedging: human judgement still vital
At the EuroFinance conference, AI also came up on the topic of managing FX operations—but always with human intervention. Nita Baindur, Associate VP, Assistant Treasurer at Agilent, stressed how AI can help map exposures more systematically, propose hedges and detect anomalies. “There’s human judgment involved in deciding how much to hedge,” she said, adding that: “Even if AI gives you the answer, you are finally accountable.”
“AI is not replacing us” – and why that matters for operating models
The title of a workshop stated clearly that: “AI is not replacing us.” As Garima Thakur, Treasurer at Creative Artist Agency noted, the real advantage of AI lies in the speedup of long processes. Her conclusion was – also – that treasurers will still be making the decisions, but their skill mix is changing
The workshop entitled “Coding for treasurers: prompt engineering for efficiency” highlighted handson ways to improve data processing and interoperability between information systems. Mario Del Natale, Treasury Director, Global Digital Treasury at Johnson Controls, showed how precise prompts and lightweight code can streamline reporting and the handling of data. His take? Overstretched IT teams cannot process every request from the treasury team quickly. Treasurers who can code—supported constructively by IT—will therefore help create value faster.
Bots and automation
Robotic Process Automation (RPA) is adapting to the advent AI. Experts on the panel “Treasury process automation: the evolution of RPAs with AI” panel discussed how “intelligent” bots can now handle more complex flows, interoperate via APIs and reduce reconciliation issues across systems. Speakers from Booking Holdings and BAT described how modern RPA can help bridge interoperability gaps across the information systems used by finance leadership, while unanimously emphasizing the value of a broader data and systems strategy rather than a simple corrective for flawed inputs.
A bot can either be a tool to speed up a process or a sign that upstream data needs fixing. In a quick survey during the stream, a majority of treasury teams said there is still no formal automation of any of their activities, while around a third (32%) said they use RPA bots and a smaller share (10%) said they already combine RPA with AI.
APIs: harder to implement than the brochures suggest
Without structured data on hand, even the best model is bound to fail. The panel “Advancing APIs: nextlevel interoperability and automation” brought together representatives from Bolt and Groupe Legris treasuries alongside provider Kyriba. The panelists explained how APIs reduce manual work and strengthen groupwide cash visibility, while noting that there is often a stubborn “last process” at the ERP or TMS level that is difficult to automate.
The session also hinted at API fatigue among a number of treasurers. BAT in particular indicated that it had stopped its projects after encountering cost, complexity and uncertain ROI. In the crosshairs: inconsistent interoperability standards and persistent dataaggregation issues that transform what should be “plugandplay” solutions into months of internal work, at the expense of higherpriority matters. A pragmatic path forward is to use prebuilt connectors where they exist and to focus on highvalue, lowcomplexity projects—in short, to manage APIs as a project portfolio with clear stage gates and kill-criteria.
Data lakes and visualization
The “Data lakes: creating and connecting data to treasury systems” session explored how centralizing bank, ERP, TMS, market and portal data helps build consistent datasets for forecasting, liquidity planning and risk analytics. It represents a first step for AI to help in identifying and preparing those datasets for downstream models. Success in this area is less about using a platform from a widely recognized brand than about governance. Crossfunctional work by IT and accounting departments is key for the success of such projects.
Once the data is trustworthy and combined, visualization becomes a lever of success. The “Show me the data again—visualisation tools for treasurers” session underlined how interactive dashboards, charts and narratives help move treasury from reporting to insight. This is where generative models are already proving useful by summarizing movements, highlighting anomalies, drafting commentary and preparing executiveready views, provided the underlying data is complete and up to date.
Governance, risk and the AI question
Treasurers repeatedly emphasized the importance of data privacy, security and data analysis. The “AI is not replacing us” piece captured the dynamic well: while some companies are testing inhouse models to keep control of sensitive data, others are using external models but with strict approvals and strategic controls in place. In every case, legal, IT security and treasury teams sit around the same table. The model that will fit your environment will depend on your risk appetite, your regulatory framework and the maturity of your data architecture.