Published: September 11, 2023
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.
AI is transforming corporate treasury and enabling data-driven decision-making. From cash forecasting to risk management, AI-powered solutions are reshaping treasury practices, empowering organisations to drive growth, and optimise working capital. However, ethical considerations and responsible governance remain crucial in leveraging AI’s potential as it transforms roles within treasury.
In 2023, AI has truly captured the zeitgeist, driven by the advance of Generative Pre-trained Transformers (GPT), in particular, OpenAI’s ChatGPT. The ethics around AI have been questioned, with the ‘godfather of AI’ Geoffrey Hinton quitting Google in May with a warning about the potential for the technology to be exploited by bad actors.[1] In the same month, Sundar Pichai, CEO of Google and Alphabet, commented that the firm was at “an exciting inflection point” as an AI-first company.[2]
AI tools are already being used in various industries, from assisting with drug discovery in healthcare to guiding autonomous vehicles in transportation, helping with process automation in manufacturing and optimising energy grids in the utilities sector.
Speaking to TMI off the record, an academic comments: “We are surrounded by AI, perhaps more so than people are aware of. That’s both fascinating and occasionally scary because some of the models have not yet matured enough to the point where they need to be, and yet they’re still deployed on various fronts. That’s the scary part, but there are some cool ideas that are safe, and that’s really exciting.”
With all the noise around the topic of AI, understanding the scope of the technology and what its applications are today, and could be in the future, is vital.
Mark D. McDonald, Senior Director, Analyst, Gartner Finance Practice, says: “People make the leap and assume that AI can do everything that humans can do, but this is totally wrong. What’s the difference between this human-like capability that AI has and being human? That’s a grey area which is creating many challenges at the moment.”
One particular area of concern is the impact AI could have on jobs. For example, one of the key pillars of this year’s strike action by the Writers Guild of America that has brought all Hollywood film and TV productions to a halt is to protect writers’ livelihoods from the perceived AI threat.[3]
Bob Stark, Head of Market Strategy, Kyriba, notes: “Six months ago, I could quite easily stand on a stage and say, ‘AI is going to complement what you do, it’s not going to replace what you do.’ Now [we have] the rise of ChatGPT and the cottage industry of generative AI around that. While it doesn’t completely change that point, it provides an avenue of extreme levels of automation that AI can provide, more than we were thinking a year ago.”
In some ways, AI may be used as a double-edged sword by business management. It can either replace a number of jobs or it can help many people do their jobs more efficiently, faster, and at scale.
“Jobs that AI bots can do today are likely going to be erased by the technology, so that is a concern to me,” reflects the academic. “But if different people, in whatever job they do, can jump onto this fast-moving train, they can gain a lot out of it as well. If each individual is able to use this technology for their own advantage or scale and become more familiarised with the use of this technology, they can also benefit. It depends if you want to be on the fast-moving train or if you’re just watching it go by.”
Stark highlights that, with the pace of change in the AI world, treasurers need to understand very quickly what impact the technology will have on their roles in the not-too-distant future.
“Some treasury professionals have to get used to the fact that – while they will not be completely automated out – the composition of their day will be different,” Stark predicts. “There’s a lot that AI will be able to accomplish in what people are currently carrying out during their daily task management.”
That is not to say that the handover of tasks to AI and a revised treasurer’s role will be a seamless change. There may well be some bumps in the road as people learn what AI can and cannot achieve.
“AI is like a new kind of co-worker, it’s a new employee in our organisation … but it’s not like the old employee,” remarks McDonald. “It has different strengths and introduces a bunch of new weaknesses. We have to learn what that new worker does.”
The ability to efficiently use cash rests, to a large extent, on a treasurer’s ability to predict organisational cash flow. Looking at the specific treasury processes that AI could disrupt, it is hard to ignore cash flow forecasting as a prime use case.
“We’re doing that with the use of the AI that we’ve built into our product for tasks such as market scenario projections,” the academic says. “These projections, together with the cash flow engine that we’ve built internally, as well as some other calculators and tools, all really help us with cash flow projections and how those financial strategy recommendations over time can increase the amount of portfolio return for the clients.”
The more accurately that a corporate can predict what its cash surplus or cash requirement will be further out into the future, the more the prediction will translate quite dramatically to the bottom line. This is an area where AI can add an extra layer of efficiency, as McDonald explains through a use case.
“If a company has to borrow money, the ability to increase the accuracy of its cash requirement – especially with high interest rates – is an advantage, even if the forecast is just two days further out than the treasurer can do right now,” he comments. “Organisations can save a lot of money on short-term borrowing costs. This is achievable and can be done today.”
Corporates have literally thousands of customers and multiple times that number of invoices to deal with. In order to simplify that, companies have standardised payment terms and conditions, and often also try to have some standardised, fixed credit terms. This is another area that AI could automate.
“Organisations may carry out a one-time credit assessment of a customer, but they might not always update that as it’s time consuming,” continues McDonald. “However, the cash positions of customers change, they are fluid and can fluctuate from quarter to quarter in many cases. We can employ our AI co-worker to help with that. That’s something many companies are not even doing now, so it’s not going to put anybody out of work, it’s going to help somebody do their job better.”
This would enable treasurers to tailor payment terms that match the customer’s needs closely to their own and increase the ability to make a sale, but also maximise their company’s ability to collect cash.
“It might get a little confusing, but we could do that on an individual invoice level, where each invoice has a different payment term,” adds McDonald.
Using AI on credit terms can enable treasurers to be more confident of the risk of that relationship. This is no longer just an intuitive assessment. Treasurers can obtain a concrete number on what that risk is. This also then enables treasurers to release trapped cash out of their cash flow.
“By using these algorithms to help maximise cash flow, treasurers can start to understand the precise elements about their businesses that are driving trapped cash,” outlines McDonald. “That lets the treasury make changes to the business to release it. That’s a little bit more advanced, but it’s a natural next step after a treasurer has been able to forecast and predict their cash.”
Perhaps no AI development has made more of an impact on the public consciousness than ChatGPT from OpenAI. This most popular example of generative AI has fired up people’s imaginations, treasurers included, as a practical example of how AI might operate.
“If you think of what GPT does right now outside of treasury, it’s very obvious as to what it’s going to look like within treasury,” comments Stark. “Think about cash forecasting. To do this without AI, treasurers get into their ERP, look at the AP data, decide what is good and what is not good, then copy and paste it into Excel. Then they’ll manipulate and play with the data because they need to organise it in a certain way, based on what they want to see in their global cash forecast, and then they need to map it into the forecast. ChatGPT can do all that.”
While a lot of that process is rules-based, the part that’s AI-oriented is around the massaging of that forecast. That’s the element that a human was doing, but it is possible to teach generative AI how the treasurer would do it, so that the AI can do it better and faster.
“What generative AI has taught us is that there’s a level of automation where we can have the AI model do all that for us, it can write its own script, as an example,” Stark adds. “If we think of every use case we had for RPA, that is completely replaced or made obsolete by what GPT can do, in addition to the learning and intelligence that it brings to doing something that’s beyond just rules-based programmatic.”
Fraud management is another prime area for GPT technology to be deployed by treasurers, particularly as the technology is also being weaponised by fraudsters on the other side of the equation. There are more linguistically complex fraud emails swirling around in the ether, written by ChatGPT. The characteristics that made it obvious to spot phishing emails are being replaced by more naturally written emails that do not appear fake at first glance.
Stark notes: “Many of the latest fraud schemes are enriched by data to create a more complex attack, but AI can ensure the treasurer identifies which payments are exceptional. Machine learning offers data-driven protection, in real-time, operating at machine speed.”
There will be additional levels to an AI security program as the technology becomes more scalable with yet more payments, with the tool providing also secondary screening.
“Ultimately, in the future, treasurers will be able to use AI to screen and validate master data changes in real-time, comparing against internal patterns and third-party data sets,” adds Stark. “That is not completely automated today, but the next generation of AI will make this process completely autonomous.”
So, while in its simplest form, generative AI tools may look like a chatbot, it’s actually a chatbot that can evolve into carrying out tasks far beyond just writing prose. It marries a process with a programmatic side.
“Our product teams are evaluating a variety of use cases in Kyriba’s innovation lab,” reveals Stark. “We are finding ChatGPT can replace what you would normally click on with a mouse, such as ‘show me my cash forecast’. And more interaction is possible, where the user would ask the TMS questions such as ‘show me what’s wrong with my cash forecast’. Integrated into a treasury platform, ChatGPT would enable treasurers to enrich forecasts, execute more advanced risk policies, and confirm the optimal level of liquidity to support free cash flow targets, EPS [earnings per share] projections and other financial KPIs.”
The academic that spoke to TMI has first-hand experience of building an AI tool for financial services.
“The product I’ve built helps with tax optimisation for retirees,” she states. “It’s an extremely complicated domain and this was my first exposure to building a financial product.”
AI is relatively new to financial companies, and there’s a really high bar for performance and the accuracy of the data for the recommendations that are provided to the clients.
“When I first started, I could see that there was not a great appetite for accepting that AI can help with writing recommendations,” she reflects. “What helped is that, since then, people have seen how AI can dramatically shorten the time it takes to compile data.”
Previously, if financial advisers were to provide what the AI product offers to clients, they would have had to spend weeks, perhaps months, considering the actuarial tables, the market scenarios, and so much more.
“This is for only one client, but by using AI in the product we’ve built, this is completely taken care of within a matter of seconds,” the academic enthuses.
As a result, financial advisers at the firm the academic works at have become extremely interested in the product.
“Many advisers were really keen on getting on-board at the pilot phase,” she reveals. “We’ve been very selective with an iterative launch of the product to a small population first and then scaling up. Our product helps to ensure that people are executing on financial strategies that give the highest return over time and paying less in taxes over the long term. It’s all about optimising the return.”
The potential of AI is immense and it will change our world. We’re standing on the brink of a massive change in the way things are going to be done, not just in treasury and finance, but in all walks of life.
“AI is like giving someone who’s blind the ability to see – and I can definitely relate to that because I was close to being blind and then underwent massive eye surgery which helped me to see again,” reveals the academic. “The same goes for people in the treasury domain, they might not know what is possible. The potential of AI is fascinating to me, and if they were to see what I see, that may well change the focus of their research or where they invest their budget. The use cases for treasury and other domains are exciting.”
There will be changes to the workforce, but we all have a choice, as McDonald notes:
“We can learn how to leverage AI, how to make it part of our lives, how to help make ourselves better or help us make our jobs better. We can also ignore it and fight it by choice, but those who do are probably going to lose the fight. Anybody who’s afraid for their career should consider that.”
Understanding AI and learning how to take advantage of it is not likely to happen organically or come without any effort. People will have to go out of their way to understand it and we will all need some new skills. Above all, understanding what the grey area is between human and human-like, and what we can rely on from new AI solutions, is essential.
“We need to understand what we can expect from AI tools but also very clearly understand where their limitations are and when we, as individuals and professionals, have to make sure that we stay engaged in the process,” McDonald adds.
The level of automation, simplicity, and even democratisation of AI has changed dramatically in 2023 alone. ChatGPT, in particular, has opened people’s eyes to a new stage of what’s possible.
“I would encourage anyone in treasury to try to follow what’s happening, look at different examples, and start asking, ‘how could that work for us?’ concludes Stark. “Being open minded to AI will enable your job – and you in it – to evolve. It’s about harnessing new data and greater automation to advance your opportunity within your organisations.”
A Beginner’s Guide to the Automation Game Automation offers treasurers a chance to make processes more efficient, save time and money, and enhance controls around cash management. For those just embarking on such a journey, understanding where to start and defining the goals is essential for a successful outcome. In today’s rapidly evolving financial landscape, […]
