Like many companies around the world, Modigent, the Pheonix, Arizona-based energy infrastructure group is gradually integrating AI into working capital management. The company’s treasury department is an early adopter of AI-powered solutions supporting cash flow visibility where Nini Johnston, Vice President of Treasury, has integrated Stampli software to help streamline invoice management in the company’s accounts payable processes. Elsewhere, she is testing Kyriba for treasury management and Quadient for receivables automation.
“Cash management and cash forecasting are at the core of what I do today and really, every tool we’re evaluating for working capital and treasury management has AI capabilities,” enthuses Johnston who joined Modigent two years ago to run treasury on her own. Today she leads a team of 20 people across working capital, risk mitigation, and treasury forecasting and analytics in reflection of the pace of growth at the company. It also speaks to the firm’s demand for working capital management across the enterprise to support its rapid expansion.
Johnston believes that one of the most compelling aspects of AI is scenario modelling and analysis, something that normally take days, or even weeks, to carry out. “These tools allow us to process data while we’re sleeping and give us the ability to focus on exceptions rather than routine processing. This means we can better quantify issues and delays instead of just guessing at their causes,” she says.
Fine-tuning predictions for the future with AI is an aspect of the job she particularly enjoys. “The forecasting aspect really excites me. Getting to see where we’re going, making our predictions more accurate, and then analysing why we missed when we do, it’s just fun. There’s something thrilling about trying to predict the future, even though you know you can’t get it exactly right.”
Getting the use case right
Modigent’s application of AI to support working capital management in a clear use case is an example of how companies are beginning to integrate the technology in a way that goes beyond using it to draft policy documents or embed it into Microsoft applications with copilot. It is also an encouraging case study at a time the hype around the technology and the overwhelming availability of AI tools from service providers has made some treasurers hesitant about its true value. Istvan Bodo, Director of Strategy and Operations at The Hackett Group, the Gen AI strategic consulting and executive advisory firm, also argues working capital management support offers the most compelling use case for AI in treasury to date.
Integrating the technology to predict future customer payments so that treasury can forecast an increase in cash flows directly informs efficiency and credit management processes, for example. Elsewhere, AI’s predictive models fed on past data visibly improve credit scoring, flags the probability of payment delays and supports inventory management. He believes routine and repetitive tasks around working capital management will be the first beneficiaries of AI which will increasingly drive new treasury operating systems boosting efficiency, accuracy and data-supported insights.
Identifying a clear use case for AI also helps counter common internal caution and resistance to the technology, continues Bodo. Articulation of the benefits like time saved or insights gained helps bring round employees who question the benefits. Resistance doesn’t necessarily stem from anxiety about job losses. It is more rooted in concerns about whether the data is reliable enough to shape important decisions. “It is really a mixture of resistance and caution at the same time,” he says.
A clear use case also supports budget allocation and helps ensure treasury has the right talent in place and data quality before pressing the AI button. Treasury shouldn’t begin integrating AI until “essential pillars” like robust data, governance, talent and internal infrastructure are safely in situ to successfully extract value from the investment, he advises.
A relevant use case also eliminates the risk of launching a large-scale transformation that can quickly become unwieldy. It focuses treasury on determining the benefit and return on the investment. “The most common question we hear is: What are the use cases that are most applicable to my process and will add most value?”
Getting the data right
It is easy to quickly run into trouble integrating AI if a company doesn’t have good customer data, reiterate Treasury Today interviewees. If organisations lack good quality data used to train the AI models, it will impact the quality of the AI predictions and recommendations which could be wrong or unreliable.
A key area of concern is inherited bias in the algorithms, and inaccurate outcomes which arise if the algorithms are trained on poor data, continues Bodo. In an added challenge, complex AI models also lack transparency. This makes it challenging to understand the rationale behind the predictions, also raising resistance regarding their reliability. “Organisations should keep in mind the risks of poor data,” he says.
“As an emerging technology, AI hasn’t been without its issues, particularly when it comes to biases in financial data. Accurate and consistent data is critical for AI to work effectively, but maintaining this in real-time financial environments is no small feat. In the treasury function, if implemented wrongly, it can lead to flawed predictions and misinformed decisions,” adds Laurent Descout, Co-founder and CEO of Neo, the FX corporate risk management and collection and payments platform.
The importance of strong data is central to strategy at Kyriba which has just rolled out a new system to enable easier access to its APIs, the critical technology that connects software applications. The company’s first-of-its-kind GenAI-driven platform, App Studio, bundles APIs into a catalogue putting data connection on a shelf where whole finance teams can access them to solve liquidity bottlenecks and offer valuable insights.
The App Studio is an example of how AI is changing treasury, Thomas Gavaghan, VP of Global Presales at Kyriba tells Treasury Today. He also flags that his conversations with treasurers about AI are still typically at a high level. Outside IT, few people understand the intricacies and variability that AI offers. “The reality is that AI is the future, and we need to address it,” he says. “The conversations we have around AI today are like APIs seven years ago.”
He agrees that one of the most useful tools emerging from AI sits around extrapolating forecasts based on historical data. Kyriba is currently extracting around 60% more data this year compared to last year. Moreover, today’s data is core, structural data spanning bank statements and historical financial transactions like FX and working capital invoices. Ensuring the data is timely, searchable and in a centralised location, is not an easy task.
“Kyriba is investing in data modelling that will build the next level of technology. This is the complex part of AI; this is the hard part,” he continues. “To be able to take in data in a matter of seconds is not trivial but this infrastructure will give us the ability to grow.”
Gavaghan observes that AI is not only being used to improve forecasting. Another area it is also changing is investment strategy. But this is an area many treasurers remain instinctively cautious of AI’s presence. Common concerns include algorithms acting outside the investment policy, for example.
“The same problem could just as easily occur with a new employee,” he reasons. “People are uncomfortable about this, but we have to start thinking this way because this is where the technology is taking us.”
Getting the talent right
Treasury teams seeking to integrate AI also need to ensure they get the talent right. Experts flag the importance of employees with data analysis and visualisation skills, equipped to interpret and turn the data into insights.
The human contribution to the treasury function in an AI world is examined by Citi analysts in the bank’s recent paper Citi GPS: Treasury 2030. The authors reflect that although human intervention will remain a core element of treasury processes in the future, humans and machines will be “more closely coupled” as innovation takes hold. Because technology can evolve faster than people, treasury needs to ensure guard rails are in place to ensure the responsible and safe use of those technologies. The authors also argue that people-friendly and hybrid working models will be important in attracting the necessary talent to support the future treasury function.
At Modigent, the shortage of talent working in tech is a key issue. Something Johnston links to the fact many graduates who originally choose finance often end up working in tech. The lack of talent is now fuelling the company’s investment in tech to reduce its reliance on people. “Finding seasoned treasury personnel for a business our size is challenging, so we’ve turned to AI and automation tools to help bridge that gap. Automation and AI represent such an important frontier for practitioners like me,” she reflects. “We need technological resources to help us continue functioning effectively since we can’t rely on growing our human resources.”
Others warn that the replacement of people with technology holds real risks. Sectors at risk include auditors and accounting firms, argues Graham Clark, COO & Co-Founder at Inflo, the digital audit platform, who says that clients value the judgment, expertise and experience of auditors, which AI cannot replicate. “Trust in financial audits comes from human interaction, not algorithms,” he says. “Human intelligence remains essential to auditing, as it provides insight, context and judgment that AI cannot match.” Achieving the correct balance between human interaction and maturing technologies requires careful thought in the payments industry, reflects Descout.
The benefits of AI in payments are already visible. The technology is improving the analytic capabilities of payment firms, meaning thousands of transactions can be checked in a couple of seconds and because AI has the ability to learn from previous payments, it can better identify cases of potential fraud.
But he argues banks should be aware of the challenges that exist around the technology before committing to further adoption. Many financial institutions rely on legacy systems that were never designed to accommodate AI and incorporating these technologies demands not just resources but also a strategic approach to minimise disruption.
“Look at 2010 when banks spent huge amounts to cope with the first wave of fintech innovation, which didn’t exactly work out for them,” he concludes. “Given banks are risk-averse institutions, there are also plenty of challenges around AI that need to be thoroughly examined first, such as data protection, before banks commit to further AI adoption in 2025.”
