Mission Intelligence

Published: September 20, 2023

Download this articles as a PDF
Mission Intelligence

Transforming Treasury Risk Management Through AI

This article forms part of TMI’s new Guide to AI in Treasury. Sign up here to receive your complimentary copy once it is published.

It is vital that AI serves as a stepping stone in risk management rather than a hindrance to organisational processes. Here, two industry experts offer an in-depth examination of the technology’s potential uses.

The saying goes that no risk means no gain. However, individuals, with their intelligence, experience, and ability to interpret data to a large extent, have managed to fine-tune the risk to extract the maximum risk-reward benefit.

Risk management in the context of the treasury department involves implementing the most appropriate measures to mitigate risk, often requiring a ‘middle-ground’ approach. Being excessively pessimistic and conservative can diminish the desired benefits or returns. But it is the job of the treasurer to preserve cash. And often, that means being conservative. So there is a balance to be struck.

Using AI to even out the equation

With the advent of microprocessors capable of processing large amounts of unstructured and de-normalised data, humans have integrated AI into the risk management equation, enabling faster, more meaningful, and more accurate outcomes.

The emergence of AI disciplines such as Generative AI (ChatGPT –  Generative Pre-trained Transformer), ML, large language model (LLM), robotic process automation (RPA), and pattern recognition is predicted to have a profound impact on the efficacy of risk management (and other treasury processes).

Broadly speaking, a treasury risk management framework can be categorised into:

  • Risk identification and assessment
  • Risk measurement
  • Risk mitigation
  • Risk monitoring and reporting
  • Risk governance

AI can be introduced into the realm of risk management, taking into careful consideration various aspects including societal dynamics, ethical mindsets, appropriate regulations, and dependable controls. As treasury endeavours to maximise investment returns while minimising costs, and ensuring the preservation of the investment principal, the incorporation of AI can significantly enhance the accuracy of projected calculations.

Risk protocol

Risk management within the treasury domain, as highlighted by the Basel Committee, the Federal Reserve, and other central banks, places significant emphasis on the primary risk framework. Diverse risks –  including credit, liquidity, market and event, sovereign, operational, currency, and reputational risks –  are considered of utmost importance.

With the inherent concepts and advantages of AI, it can be utilised to identify associated risks and their correlations at specified target areas. The computational capabilities of decision tree models and their proficiency in solving complex problems using datasets make AI a transformative force in the continuous pursuit of risk control. AIs LLM operates on the basic principles of summarisation, inference, transformation, and reinforcement. (See diagram below.)



The central idea, and indeed the primary driving factor for embracing AI in treasury risk management, is not only to detect risks associated with various treasury functions at an early stage but also to predict them with a high degree of accuracy in advance. Consequently, informed decisions can be made and necessary corrective actions can be taken. The exceptional data analytics and computational capabilities of AI greatly facilitate this task.

Susceptibilities and consequences

Within treasury the prevalent risks usually revolve around liquidity, market fluctuations, operational challenges, and counterparty risk. The following activities are particularly susceptible to risk exposure and its subsequent consequences:

Liquidity risk with cash flow forecasting

Cash flow forecasting has traditionally been a key focal point in treasury operations, involving the thoughtful consideration of intelligent data and informed decision-. AI has the potential to delve deeply into this activity, enhancing liquidity management to a great extent. It enables treasurers to make more effective interim and strategic cash flow decisions, empowering them to implement timely provisional and remedial asset-liability measures.

Stress testing:

AI has the capability to simulate various market scenarios and conduct stress tests to assess the impact of different events and activities on cash flows. By leveraging analytical probability derived from extensive historical Big Data and employing models for extreme case impacts, AI enables more accurate predictions of extreme scenarios, thus improving preparedness for outliers. By providing specific use cases that can be applied to models and algorithms, AI assists treasury teams in understanding potential risks and identifying vulnerabilities that have been addressed through strategic workarounds. This facilitates adjustments to their strategies, resulting in enhanced risk management and financial planning.

Payments due diligence

Leveraging advanced algorithms, AI-based sanctions screening in treasury payments detects politically exposed persons (PEPs) and restricted countries, safeguarding against illicit transactions and maintaining legal cash flow within an organisation. This ensures authenticity in payment processes.

Investment portfolio monitoring

Artificial intelligence empowers corporate treasurers to efficiently manage their portfolios by automating data analysis, performance tracking, and trend analysis, enabling well-informed decisions for optimised re-balancing and deployment of the most favourable model.

Fraud detection and AML monitoring

Leveraging AI algorithms, transactional data can be meticulously examined to uncover anomalies or peculiar patterns that may indicate the presence of fraudulent activities. As fraudsters continually adapt and employ innovative methods, such as manipulating funds and concealing counterparties to elude traditional fraud detection measures, AI's analytical prowess plays a vital role in efficiently and effectively detecting such occurrences. Many banks will offer such services to their corporate clients, so treasurers can indirectly leverage these benefits. But there is also a growing number of third-party services available for treasurers to leverage. But that requires investment dollars and not all corporate organisations are yet at the stage of allocating this kind of budget to fighting fraud.

Risk measurement and automated monitoring:

Treasury teams use Cash Flow at Risk (CFaR), a financial metric, to evaluate how adverse events or market changes might impact a company's cash flow, enabling proactive strategies for liquidity and financial stability. AI-powered systems enhance risk prediction since they can can autonomously identify and monitor a range of risk factors, encompassing credit risk, market risk, liquidity risk, and operational risk. These systems can continuously observe key indicators and generate alerts or reports whenever risk thresholds are surpassed, empowering treasury teams to take prompt action.

Avoid non-compliance of regulatory reporting

AI has the potential to automate compliance processes and fulfil regulatory reporting requirements. By deciphering complex regulations, AI systems can ensure that treasury operations adhere to relevant guidelines, reducing the need for manual effort and minimising the risk of non-compliance. Through prompt engineering and user feedback, these systems can automatically generate user-defined narratives, ensuring timely and compliant practices are followed.

AI in action

All of the use cases above can be executed using a  fine-tuned LLM model such as GPT3/BERT (see diagram below) which is trained using structured or semi-structured data, achieved through a parser. When prompts are executed, responses are generated by conducting content analysis through prompt engineering. The output comprises natural language text that includes analysis results such as summarisation, inference, and expansion.


Base LLM Coupled with Instruction Tuned LLM

  • Input Source: Transactional data (JSON)
  • LLM Model: GPT3/BERT
  • Query enhancement/Creation through reinforcement learning from human feedback (RLHF)

Conceptual Framework

Weighing the pros and cons

Regardless of the many potential applications of AI, it is paramount that this technology serves as a stepping stone in risk management rather than a hindrance to organisational processes. Furthermore, it is crucial that AI acts as a supportive element in risk mitigation and does not become an additional dimension that could undermine the treasury function.

Sudeep Mukherjee is a senior consulting professional working in a renowned technology MNC. He has collaborated with global financial services sector firms worldwide, encompassing diverse projects including technology-enabled digital capabilities, legacy system and process modernisation and business-driven IT operations. Mukherjee holds Financial Risk Manager (FRM) certification from the Global Association of Risk Professionals (GARP) and has a Master’s degree in business management with a specialisation in finance, complemented by his Bachelor’s degree in computer science.

Raja Basu is a Senior Consulting professional in a leading technology MNC. In his current role he works as a business architect and helps clients to realise their digital transformation journey. He has special interest in responsible use of AI and is also pursuing his PhD from the Xavier School of Management, Jamshedpur, India.

Sign up for free to read the full article

Download this articles as a PDF
Article Last Updated: May 22, 2024

Related Content