Everyone’s talking about the potential AI has to change the world. But what can it do for treasurers?
As Royston Da Costa, assistant treasurer at Ferguson, points out: “ChatGPT has democratised artificial intelligence, making it easier to use for non-technical staff. This has opened up new opportunities for adoption across treasury.”
According to Da Costa, the possible applications of AI in treasury include reconciliation, forecasting, and optimisation of actions such as hedging, investing, borrowing and trading. “AI can use optimisation (genetic) algorithms that consider multiple objectives, constraints, and risks to find the best solutions,” he adds.
Where fraud prevention is concerned, AI has much to offer in detecting and preventing fraud in payments, invoices and transactions, as well as providing insights and suggestions for improving trading performance. Da Costa also notes that AI can help execute intelligent trades in foreign exchange, bonds and other markets, using machine learning and natural language processing to analyse market trends, predictions and liquidity needs.
ChatGPT has opened up new opportunities for adoption across treasury
For James Kelly, senior vice president of treasury, risk management and insurance at Pearson, harnessing AI is already a reality. In 2018, the company began implementing an AI-powered cash forecasting solution that has yielded numerous benefits, from improving working capital metrics to reducing borrowing requirements.
“When you say ‘AI and cash forecasting’ in a single sentence, it sounds like it must be straightforward,” comments Kelly. “But there’s quite a lot to it.” For one thing, he says, the company uses AI for categorisation of data and is working towards using machine learning to recommend new categories.
“Then, in terms of generating new forecasts, we use it in cases where the data is relatively regular. So, you can take patterns either from past performance or trend analysis, with information about what the inputs are going forward,” Kelly notes.
Data visualisation tools make it easy to see the key drivers of changes, which is particularly useful for actual versus forecast work. The timing of cash flows can also be predicted more accurately. As Kelly explains, knowing whether the company is ahead or behind the forecast at a particular moment in time isn’t necessarily useful in itself, “because it doesn’t tell you whether your view of the year end position is correct. It’s about validating not just what’s happened, but why, and what that means for your expectations.”
High quality outputs can only be generated through high quality inputs
But while AI opens up numerous opportunities for improvement, treasurers may need to overcome various obstacles along the way, including algorithmic limitations, incomplete data, resistance to change within the organisation, and challenges around the ability to interpret results.
“As in any technological implementation, it is data quality and availability that is key, along with high refresh rates,” says Carl Sharman, a partner in Deloitte’s performance improvement practice. “High quality outputs can only be generated through high quality inputs, and integration with the organisation's existing data sources is critical.”
Defining return on investment and communicating the benefits to senior stakeholders presents another challenge. And while use cases may be expanding, Sharman argues that without common AI standards and renowned vendors – many are start-ups and fintechs – “the choice in being an innovator, early adopter or fast follower becomes more conservative”.
When adopting AI technology, Sharman suggests that treasurers consider starting with small, discrete pilot projects “in order to test theories and AI’s effectiveness and impact on cash management”, before implementing the technology more broadly.
Rebecca Brace is a freelance journalist