How automation and AI

Unlike the initial major revolutions in the workplace, which took years and decades to fully take hold, artificial intelligence (AI) has been making inroads into the world of work and other areas of life at a remarkable pace. Broadly speaking, AI refers to the ability of computers to solve specific tasks with "human-like" intelligence. Even today, a number of corporate departments are using tools and computers that are based on AI models. In Treasury, too, companies and system providers are already discovering initial uses for this technology.

And yet the dynamism and speed with which AI solutions are transforming a wide variety of business processes are only perceptible to a limited extent in treasury departments. At the moment, it is unclear where the technologies can be used, and for what purposes they can be used remains an open question. According to a recent study conducted by HSBC in cooperation with DerTreasurer, as many as one-third of all treasury departments have not even begun to explore the new technologies and possibilities¹. So it is high time to take a detailed look at the potentials, but also the challenges and dangers of using AI solutions for treasury departments.

Virtually no other topic is causing as much concern among companies of all sizes as the concept of artificial intelligence - or AI for short. Just explaining its underlying logic and distinguishing it from related terms could probably fill several books these days. At this point, we will therefore deliberately refrain from giving a detailed technical definition. In everyday language in particular, the elasticity of the word AI in terms of its abbreviations and terms is very high. Still, in order to be able to evaluate its potential and dangers for Treasury, it is important to at least provide a rough definition of the central terms in the AI universe. 

In technical terms, artificial intelligence is a subfield of computer science and is the concept of transferring the problem-solving and decision-making abilities of humans (i.e., human intelligence) to computers and machines. As opposed to well-known automation solutions (such as those used in Robotic Process Automation (RPA)), in which a machine is given individual steps and processes them in a specific order, AI-based tools are based on algorithms. As a result, such tools are capable of independently adapting decision parameters to a specific problem, for example. At its core, AI attempts to find common patterns in the data to be analyzed, to structure them, and then to derive rules using its own resources. This results in decision trees that the AI uses to continuously learn, optimize and ultimately solve even complex problems in a "human-like" manner.

Contrary to popular belief, the term AI (= Artificial Intelligence) is not a clearly distinguishable concept, but merely a semantic one. The term machine learning is also frequently mentioned in connection with AI. It can basically be defined as a subset of the AI domain. 

While the term AI defines the overarching concept for how a system or machine "works”, machine learning refers to a set of methods that are used for developing those systems or machines - in other words, a way of implementing AI. As the term already suggests, these methods have the learning process of the systems in focus. In designing such methods, no clear programming commands are issued, but rather the system is made capable of "learning" with the help of algorithms, huge amounts of data and insights.

In contrast to AI programs, automation solutions have already been in use in treasury departments for quite some time. Examples include payment transactions and trade finance, but in practice such solutions are used in many more areas of Treasury.

Payment transactions 
Already today, payment instructions are frequently automated in payment transactions for a more efficient and faster payment process. As a result of automation, payments go through faster, with less manual errors and greater accuracy. Also, automated email notifications can be sent to any relevant staff to communicate treasury actions or upcoming deadlines. This enables them to respond quickly to important upcoming key dates, such as payment deadlines, and ensure that all necessary actions are taken in a timely manner.

Trade Finance
In the trade finance sector, too, a number of system solutions have been developed to automate and standardize processes. There are many companies, for instance, that use automated workflows for their guarantee lifecycle management. In addition to reducing manual errors and increasing accuracy, automation also helps meet compliance requirements, as all processes are standardized and documented. 

Automation potential of the TMS
There is no compelling need for separate system solutions when it comes to using automation. The treasury management systems already in place in many companies today also permit the automation of selected processes. It is recommended that treasurers analyze these potentials in detail so they can make the best possible use of the potential offered by the system solutions that are already in place. Overall, automation in Treasury can help streamline processes while minimizing risks.

On top of the automation solutions already established in Treasury, a growing number of AI solutions are also being used. AI solutions can do more than just help make processes more efficient as well, they can also actively assist in decision-making. Based on the analysis of large amounts of data, AI models can provide valuable insights and can make decisions independently based on predefined rules and algorithms. By implementing AI solutions, organizations can further streamline their processes and improve their competitive edge. A selection from the range of solutions available on the market is intended to show the potential of AI-based systems.

Cash Management
In the face of high market volatility and uncertainty, a growing number of treasurers are concerned with their company's liquidity to cope with times of crisis. Being able to predict cash flows is critical for effective financial management, informed business decisions, and long-term success. With machine learning and predictive analytics, it's easy to create automated, accurate forecasts, minimize risk, and enable optimized decision-making. Financing decisions are underpinned by scenario simulation and stress testing virtually at the touch of a button. Key treasury management system providers have already integrated AI into their cash management and liquidity planning modules. In doing so, they make use of various modeling options, such as ARIMA(X), additive regression and neural networks. ML tools can uncover complex relationships that may be overlooked by humans. That said, there is value in human expertise combined with advanced tools to optimize predictions and be prepared for unforeseen events. Making successful use of them requires clean data, correct driver selection, appropriate IT architecture, business and technical expertise, and clear communication. Beyond that, analytical expertise is critical to correctly interpret the results and translate them into sound business decisions. We have already addressed this topic extensively in our article "Ways to optimize cash forecasting with artificial intelligence (AI)" (issue 131).

Payment Fraud Detection
Applying AI to payments is another example of using machine learning to detect anomalies and prevent fraud. An increasing number of systems providers are also integrating these solutions into their treasury management solution. For instance, payment fraud detection software uses AI to check payments against historical payment data from a data source. An engine then analyzes this data, enabling it to identify anomalies in any future activity. In general, the solutions will provide users with insight into computing variables and show how payment reconciliation is calculated. It allows users to set their own tolerance levels to determine the sensitivity for fraud detection. This is important because one of the biggest challenges is to avoid having the system return too many false positives, which disrupts the payment process. With an integrated workflow in the treasury management system, any suspicious payments can first be reviewed and then either rejected or cleared for further approvals. The use of AI in payments provides many benefits, then, particularly in terms of fraud detection and improving the efficiency of the payment process.

AI Algorithmic Trading
Also outside of the treasury management system, AI solutions are increasingly used. In high-volume derivatives trading (for commodities, FX trades, or interest rate hedges, for example), they are used to suggest trading strategies and control automated trading systems. This involves adding AI components to conventional algorithmic trading systems. The term conventional algorithmic trading refers to the method of trading that uses solutions to make trading decisions in an automated manner. These programs analyze market data and make decisions based on predefined rules and algorithms with the goal of making trading faster and more efficient.

AI-powered systems can go beyond this to autonomously identify and execute trades, manage risk, and manage the flow of orders. By doing so, they can improve liquidity management and execution of large orders by dynamically optimizing size, duration, and order size based on current market conditions. Where AI-driven systems differ from conventional trading is that the AI model learns and adapts to changing market conditions. In the process, the time delay caused by human intervention is drastically reduced.

The talk of the town at the moment is not only AI in general, but also specifically the chatbot development ChatGPT, which enables text-based exchange with users by means of AI. The abbreviation "GPT" stands for "Generative Pretrained Transformer". It is a type of artificial intelligence based on the Transformer architecture and trained to understand and generate natural language ("Natural Language Processing", NLP for short). Such models can generate text, answer questions, translate text, and much more. Anyone who has already used ChatGPT can easily think of several use cases for Treasury: creating and commenting on financial reports, analyzing data, or editing and summarizing contracts. The application possibilities seem almost endless. But still: ChatGPT and similar, publicly accessible programs cannot be used in Treasury for the time being without any concerns for data protection. For this reason, a growing number of companies are developing their own secured solutions to meet these challenges. But even with such in-house and safeguarded solutions, there is still the question of whether Treasury is willing to share sensitive financial data within the company as a whole.

A potential answer to this question could be to integrate GPT models into treasury management systems. Doing so could help build a bridge between strict data protection and Treasury's need for information. Here, the idea is that its integration could help to process sensitive data without overstepping the bounds of internal security measures. Just how willing treasury management system vendors are to invest in this will point the way forward for this type of technology. In any event, it is going to be interesting to observe these developments.

Artificial Intelligence (AI) has already made inroads in Treasury and will transform and shape many aspects of the function in the future. Even now, progressing automation is making sure that repetitive tasks are carried out efficiently and quickly. AI will amplify this trend even further and set new standards in data processing and analysis. With big data being processed and, in the best case scenario, in real time, predictive capabilities will be boosted on many levels. That will lead to better informed decisions across all elements of control, such as liquidity or risk. AI-based models hold the potential to play a growing role in forecasting market developments, interest rates and currency fluctuations, as well as in risk management. Leveraging machine learning algorithms, these models can analyze vast amounts of data and recognize patterns that are difficult for human experts to detect. As a result, they can help identify risks more effectively and make informed decisions. 

However, deploying AI comes not only with opportunities, but also with risks and challenges that need to be factored into any new deployment of AI functionalities. Owing to their function as a central point of contact for financial information and key figures, treasury departments routinely collect and process highly sensitive data. For this reason, as with all new implementations in Treasury, careful attention must be paid to keeping an eye on data protection requirements. Any use of cloud solutions, which are frequently used due to the large volumes of data involved in AI technologies, should therefore be treated with caution: it should always be carefully checked where the sensitive financial and personal data is stored (i.e., where the provider's servers are geographically located) and whether the data storage meets company-specific data protection requirements. And as with any data processing, the garbage-in/garbage-out principle comes into play. AI systems too rely on valid or high-quality input data, or else they may produce erroneous analysis results.

In addition, new ethical demands are placed on Treasury due to the use of AI. Any deployment should be reviewed to determine the extent to which unintended bias or discrimination could occur. One such example is the use of AI-based models to predict credit risk, where the models can make predictions based on historical data and other factors such as demographic characteristics and credit behavior patterns. These also give rise to new legal issues in terms of liability for AI-based decisions in Treasury.

Without a doubt, our findings so far have made it clear that the increasing use of artificial intelligence will not only transform systems and processes, but also their users. Among other things, the requirements profile for a treasury manager will involve greater technological sophistication in the future. After all, the benefits and impact of AI on Treasury will need to be gauged to ensure that technologies are used in a targeted manner and support business objectives in the best possible way. Treasurers will also need to make sure that the data and executions generated are valid and reliable. Also, attention must be paid to the security and protection of AI systems from cyberattacks. In this respect, critically questioning the results of AI solutions will become an essential task for treasurers. One effective means of preparing employees for the new challenges could be targeted training in the use of AI solutions.

All the same, thanks to the increasing degree of automation, treasurers will be able to spend more time on strategic considerations and decisions. Rather than preparing data and calculating key figures, the focus will shift more to the needs of various stakeholders. The needs of customers, suppliers, banks and other business units must be recorded and taken into account - with the ultimate aim of ensuring that the company's financial positioning is in line with its business objectives. Having said that, it is vital to emphasize that despite the use of AI, data validation and verification by humans remain indispensable. Treasurers will want to be in a position to assess data accuracy and ensure that AI analyses are in line with business objectives.

As the world keeps on changing faster and faster, artificial intelligence will surely be the next big revolution. It will be imperative to consistently exploit opportunities such as efficiency gains, flexibility, and improved decision-making, both in general and in the treasury area. And yet, AI also brings challenges and risks. It requires proper use and handling, especially when it comes to sensitive data, which, among other things, is largely located within Treasury. At the panel discussion of this year's Digital Treasury Summit 2023, we will also be tackling this topic and discussing in-depth insights around the opportunities and limitations of the possible applications of AI in Treasury.

Source: KPMG Corporate Treasury News, Edition 136, September 2023
Authors:
Nils Bothe, Partner, Finance and Treasury  Management, Corporate Treasury Advisory, KPMG AG
Karin Schmidt, Senior Managerin, Finance and Treasury  Management, Corporate Treasury Advisory, KPMG AG 

_______________________________________________________________________________________________________________________

¹ DerTreasurer & HSBC, 2023: KI, Robotics & Co.: So beurteilen Treasurer die Chancen neuer Technologien [AI, Robotics & Co.: How Treasurers See the Opportunities of Emerging Technologies]