Kantox has been awarded Best Risk Management Solution 2025 by TMI
Read it here

CurrencyCast

CurrencyCast is a treasury podcast series from currency management experts. In each episode, we look at the pressing foreign exchange (FX) risk issues facing treasurers and CFOs today and help them identify the potential gaps in their FX risk management strategy.

AI, FX Narratives, ESG and Risk Management in 2025 with Eleanor Hill (Treasury Storyteller)

January 22, 2025
or watch this episode on

AI in Treasury: Beyond Cash Forecasting - Insights from Eleanor Hill

In a recent CurrencyCast episode, Eleanor Hill founder of Treasury Storyteller and Treasury journalist, challenged the conventional wisdom around artificial intelligence in Treasury management, offering a refreshing perspective on where the real value of AI lies.

The Evolution of Treasury Technology

Treasury's digital transformation has evolved significantly over the past decade. What started with the adoption of Treasury Management Systems (TMS) and the automation of basic operations has now entered a new phase with artificial intelligence. This evolution reflects Treasury's expanding role from a purely operational function to a strategic business partner.

While early automation focused on streamlining routine tasks and centralising data, today's AI-driven transformation promises to enhance decision-making capabilities and provide predictive insights. This progression comes at a crucial time, as Treasury departments face increasing pressure to do more with less while managing growing complexities in global operations.

Global Adoption Patterns

The adoption of AI in Treasury varies significantly across regions, revealing interesting patterns in digital transformation approaches. According to recent data, India leads organisational AI adoption at 59%, followed by the UAE, Singapore, and China, while Western markets generally show more conservative adoption rates.

This pattern reflects a broader trend in financial technology adoption, where emerging markets often leapfrog traditional technologies, embracing innovative solutions more readily than their Western counterparts. This phenomenon is particularly evident in regions where legacy systems are less entrenched, allowing for more agile adoption of new technologies.

Beyond the Cash Forecasting Hype

While many Treasury teams focus on AI-powered cash forecasting, as exemplified by ASML's impressive improvement from 70% to 96% accuracy, Hill advocates for a different approach. "Everyone focuses on cash forecasting... They're looking at AI as a silver bullet," she notes. Instead, she suggests exploring "lower-hanging fruit" that could potentially deliver more immediate value.

Practical AI Applications in Treasury

Hill highlighted several compelling use cases already being implemented:

  1. Fraud Detection: A UAE Treasury function built an in-house AI tool specifically for identifying fraudulent payments and catching duplicate supplier payments.
  2. KYC Automation: Former Pearson treasurer James Kelly is developing an AI tool that automatically completes KYC forms by pulling data from various sources.
  3. FX Management: AI can analyse large datasets to recommend optimised hedging strategies and improve exposure management.
  4. Treasury Assistant: AI can handle basic but time-consuming tasks like scheduling and automated email responses, effectively serving as an additional team member.
"Using AI essentially as a Treasury assistant... you can get it to schedule tasks for you, send automated replies to certain emails. It's just some super simple things that are really quick wins that essentially give you more resources within your Treasury Department." - Eleanor Hill

Implementation Considerations

The successful implementation of AI applications in Treasury requires careful consideration of four key areas:

Data Requirements

Data forms the foundation of any AI implementation in Treasury. Teams must first identify all necessary data sources, which often span multiple systems, departments, and external providers. The quality of this data is paramount, AI systems are only as good as the data they're trained on.

Treasury teams should establish robust processes for data validation, cleaning, and maintenance. Integration poses its own challenges, particularly when dealing with legacy systems or multiple data formats. Teams should plan for ongoing data governance and quality control measures to ensure their AI applications continue to perform effectively.

Resource Planning

Understanding resource requirements is crucial for successful AI implementation. Internal expertise needs should be assessed early in the process, identifying any skills gaps that need to be filled through training or external hiring. The role of vendors should be clearly defined, particularly for specialised applications or ongoing support.

Training requirements often extend beyond the technical team to include end-users who will interact with the AI systems daily. Treasury teams should develop comprehensive training programs that cover both technical aspects and practical applications, ensuring staff are comfortable and confident using new AI tools.

Measuring Success

Defining and tracking success metrics is essential for justifying AI investments and ensuring continuous improvement. Return on Investment (ROI) calculations should consider both quantitative benefits, such as time savings and error reduction, and qualitative improvements like enhanced decision-making capability and risk management.

Key Performance Indicators (KPIs) should be established before implementation begins, with clear benchmarks for measuring progress. Success metrics might include reduction in processing times, improvement in forecast accuracy, or decreased error rates in routine tasks. Regular reviews of these metrics help identify areas for optimisation and ensure the AI implementation continues to deliver value.

Risk Management Framework

A comprehensive risk management approach is essential for AI implementation in Treasury. Controls need to be carefully designed to monitor AI system performance and catch any anomalies or errors before they impact operations. Exception handling procedures should be clearly documented, with defined escalation paths and responsibility assignments.

Backup systems and contingency plans are crucial, Treasury teams need to ensure business continuity even if AI systems encounter problems. This includes maintaining manual processing capabilities and regular testing of failover procedures.

Change Management Strategy

The human element of AI implementation cannot be overlooked. A well-planned change management strategy helps ensure smooth adoption and maximises the benefits of new AI capabilities. This includes clear communication about how AI will affect daily operations, what benefits it will bring, and how roles might evolve. ç

Stakeholder engagement should begin early and continue throughout the implementation process, with regular opportunities for feedback and adjustment. Training should be ongoing, with support systems in place to help users adapt to new ways of working.

Technology Infrastructure Assessment

Before implementing AI solutions, Treasury teams must evaluate their existing technology infrastructure. This includes assessing current systems' capabilities, identifying potential integration points, and determining any necessary upgrades.

Cloud vs. on-premise solutions should be evaluated based on security requirements, data volumes, and processing needs. Teams should also consider scalability requirements and ensure their infrastructure can grow with increasing AI capabilities and data volumes.

Vendor Selection and Management

Choosing the right technology partners is crucial for AI implementation success. Teams should develop clear criteria for vendor selection, including technical capabilities, industry experience, and support infrastructure. Service Level Agreements (SLAs) should be carefully negotiated to ensure appropriate support levels and response times.

Regular vendor reviews help ensure continued alignment with Treasury needs and optimal performance of AI solutions.

The Human-AI Integration Model

Drawing from Professor Ethan Mollick's book "Co-Intelligence", Hill describes two approaches to AI integration:

  • The Centaur Model: A clear division of labor between human and AI tasks
  • The Cyborg Model: A more integrated approach where AI enhances human capabilities
"a blend of a machine and a person... it's using AI to enhance your human skills and to make you better at work." - Eleanor Hill

The key takeaway? AI implementation in Treasury isn't about replacement but enhancement. As Hill emphasises, "There's always still a human in the loop."

A Bottom-Up Approach to AI Adoption

Rather than starting with complex, transformative projects, Hill recommends beginning with smaller, manageable initiatives. This approach allows Treasury teams to:

  • Test and understand AI's limitations
  • Navigate compliance and data privacy considerations
  • Build confidence and expertise gradually
  • Achieve quick wins that demonstrate value

Looking Ahead

As AI technology continues to mature, its role in Treasury operations will likely expand. However, the key to successful implementation lies not in chasing the next big thing, but in identifying practical, value-adding applications that enhance existing processes while maintaining human oversight and strategic direction.

The message is clear: While AI presents exciting opportunities for Treasury transformation, the most successful implementations will be those that start small, focus on practical applications, and maintain a balanced approach to human-AI collaboration.

As Treasury continues its digital transformation journey, success will depend not on the technology itself, but on how well Treasury teams can integrate these new capabilities into their operations while maintaining strong governance and risk management. CFOs must have a broader digital transformation strategy that enhances Treasury's ability to support business objectives.

Check out our report to prepare your currency management roadmap on the eventful year ahead.

Keep Reading
No items found.