Getting AI right

Integrating artificial intelligence is a key to building a strategic framework for treasury functions, says PwC’s Nishchay Nagpal
 

Artificial intelligence (AI) has been a transformative force in finance for decades, evolving from automating rule-based tasks like reconciliation to enabling complex statistical modelling for market prediction, risk assessment, dynamic pricing and fraud detection. Today, as AI continues to mature, treasurers have a unique opportunity to thoughtfully integrate AI into their operations, unlocking greater efficiency and deeper insights in an increasingly data-driven world.

Previously, AI development was limited to experts in mathematics and computer science. However, the emergence of generative AI (GenAI) and large language models (LLMs) have revolutionised the field, making AI accessible to non-technical users.

This shift has sparked interest in corporate treasuries, which have traditionally been slower to adopt AI in their operations. For treasurers, it is imperative to look beyond the hype surrounding AI’s potential benefits and focus on understanding its various components and capabilities to identify the right solutions for their needs.

Understand key components

AI can simply be defined as systems designed to perform tasks typically requiring human intelligence. For treasurers, understanding AI's key components is crucial to selecting the right solution:

• Machine learning: A subset of AI that learns from historical data to identify patterns and make predictions. For example, it can help forecast cash flows based on past performance.
• Supervised learning: A form of ML that uses labelled data—such as past treasury transactions—to build predictive models for future outcomes.
• Deep learning: Advanced ML that uses neural networks to process vast amounts of data, useful for uncovering complex relationships in market trends or payment anomalies.
• GenAI: A breakthrough in AI that creates new content, like summarising treasury KPIs in simple, clear language.

Each of the AI components brings unique capabilities suited to specific tasks. Machine learning is well-suited for tasks that involve structured and data-heavy activities, such as analysing numbers to identify trends, uncover hidden patterns, and detect anomalies in datasets. On the other hand, GenAI is designed for tasks that require human-like reasoning and creativity, making it ideal for summarising complex analyses and generating natural language responses.

Some key treasury activities that align with these technologies and represent relevant use cases include:

For machine learning:

• Predicting future cash flows to improve the assessment and management of cash requirements
• Enhancing risk assessment to facilitate early detection and management of various types of financial risks
• Identifying and highlighting anomalous activities, such as fraudulent payments or erroneous transactions.

For GenAI:

• Explaining cash flow variances in straightforward language
• Summarising investment performance
• Creating summary briefs for management on treasury performance.

Identifying AI solutions

Integrating AI into treasury functions can streamline operations, enhance efficiency, and allow treasurers to focus more on strategic decision-making. But how can they identify the most suitable AI solution for their operations?

Treasurers must assess their maturity in deploying AI and identify key processes that could benefit from its implementation. A structured framework can guide this decision-making process, focusing on two critical lenses:

1. Maturity lens: Evaluate your organisation’s readiness to integrate AI into Treasury and finance processes. If AI has already been implemented in other finance functions, leverage those models while adjusting for Treasury-specific data and requirements. For organisations new to AI, starting small with basic models that categorize data or predict outcomes is more practical than attempting complex implementations.

2. Critical business process lens: Engage with your organisation’s treasury experts to identify key challenges that AI can effectively address, and collaborate with IT experts to assess the most suitable tools and strategies for achieving AI automation goals. This partnership can offer valuable insights into identifying and implementing AI-driven solutions, particularly for tasks that require high accuracy and currently involve significant manual effort.

Unlocking potential

By assessing their readiness and starting with simpler ML solutions, treasurers can unlock the potential of AI. Pairing ML for structured data analysis with GenAI for natural language outputs further enhances their ability to deliver clearer insights and drive continued innovation in Treasury operations.

Selecting the right AI solution is just the beginning. Treasurers must collaborate with experts who combine technical proficiency in AI and data analysis with an in-depth understanding of treasury functions. Such collaboration is essential to address critical next steps, including ensuring the availability of high-quality data, selecting AI models tailored to treasury-specific needs, and leveraging existing technology infrastructure to implement effective AI-driven automation.

About the author

Nishchay Nagpal is a corporate treasury adviser at PwC