Low-hanging fruit

George Dessing, Executive Vice President, Treasury & Risk at Wolters Kluwer, explains that 50% of the company’s digital revenues now come from products that leverage AI. “We see the impact these products have on our customers and are eager to leverage our company’s ‘digital DNA’ to implement AI into our own internal operations.”

At present, he says his team is still in the exploratory phase when it comes to implementing AI in treasury. “We have identified some use cases related to cash flow forecasting, for example.”

Dessing adds that he sees AI use cases as “the low-hanging fruit that will eliminate some of the ‘grunt’ work for the team.

“That way, the team can allocate more time to the treasury matters that they are passionate about, while upskilling them in the use of this technology that I believe will only become more important in the future.”

Faster, smarter decisions

Kemi Bolarin, Head of Treasury – Europe at GXO Logistics, says that for her team, AI in treasury “isn’t just a future concept – it’s an active and evolving part of our journey.”

She explains that following the company’s spin-off four years ago, “we had the unique opportunity to build our European treasury team from the ground up.” Initially, this meant taking a pragmatic approach, and keeping things simple and manual while building solid foundational processes.

With a clear view of where the bottlenecks and opportunities lay, she says the team is now moving into a transformational phase – “one where AI is positioned to play a central role in how we evolve.”

The immediate focus is on harnessing predictive analytics to improve the accuracy of cash flow forecasting. “In a fast-moving and margin-sensitive industry like logistics, where payment terms and customer behaviours can vary significantly, forecasting can be especially challenging,” she notes. “So how can we utilise AI algorithms to help us make sense of large volumes of internal and external data, uncovering trends and flagging unusual patterns that might otherwise go unnoticed?”

Other areas of exploration include the role AI can play in helping to detect payment anomalies and fraud, as well as enhancing the cash conversion cycle by leveraging AI in dynamic discounting and payment prioritisation. “The ultimate goal is to make faster, smarter treasury decisions, with fewer manual touchpoints and more confidence in the data driving those decisions,” she says.

Barriers to adoption

Despite the potential benefits, many treasury teams are at a very early stage of AI adoption. For example, PwC’s report found that only 26% of respondents described their AI capabilities as either moderately or very mature. Thirty one percent said they were in the early stages of development and implementation, while 42% were in an experimental or pilot phase.

Key blockers cited in the report cited include data quality and the limited availability of skills. And as Camerinelli points out, obstacles may also include data security concerns around feeding sensitive financial information into AI systems.

He adds that treasury teams may also be held back by a lack of proper technical expertise and data organisation, as well as integration challenges with existing fragmented technology stacks, “and audit trail difficulties due to AI’s non-deterministic nature.”

Deloitte’s Matsoukas comments that the biggest hurdles “are no longer the algorithms themselves but the plumbing and governance around them. Fragmented source systems, patchy master data and opaque models unsettle audit committees.”

Return on investment

At Wolters Kluwer, Dessing explains that innovation has always been a driving force within the company. “However, when implementing AI and making decisions on where to allocate resources you are really looking for scale to get the biggest return on your investment.”

He explains that the company’s treasury function represents around 1% of the total finance community. “It is more difficult for me to justify large investments for AI in treasury at the moment, as the return will be bigger if focused on other larger disciplines within the finance community,” he says.

“There is the clear conviction that if we do not grow on our current head start, we might even be missing the AI ‘boat’ later on.”

“The ultimate goal is to make faster, smarter treasury decisions, with fewer manual touchpoints and more confidence in the data driving those decisions.”

Kemi Bolarin, Head of Treasury – Europe, GXO Logistics

Confidence gap

On another note, Nguyen argues that many professionals see AI as overly complex or technical. “And sometimes AI is seen as a tech team responsibility – some people will say, ‘I don’t want to touch it.’ There’s a big confidence gap, but I always stress that AI education should be industry-wide and inclusive, not just for tech specialists.”

In addition, she notes that many people are worried about being replaced by AI. “But for me, it’s about how AI can augment people’s jobs, and what they are doing in their daily lives.”

To address these hurdles, CFTE has launched a course on Generative AI in Corporate Treasury, which breaks down AI for professionals and creates a learning path, “so that everyone can be engaged with AI, without fear or judgement.”

Royston Da Costa, Assistant Treasurer at Ferguson, and a senior lecturer on the CFTE course – explains that he has taken a proactive approach to educating himself on AI adoption. This includes understanding core AI concepts, reading focused industry reports and following treasury tech vendors that are using AI, as well as engaging with peers and experts in finance and AI communities.

Other steps that treasurers can take include attending treasury and finance events with an AI track and experimenting with AI-driven tools such as ChatGPT or Claude to simulate AI-generated reporting or forecasting. But as Da Costa notes, “Caution MUST be exercised when entering sensitive data into a public website like ChatGPT, as the data could be accessed globally!”

Taking the plunge

So where should treasury teams begin when seeking to adopt AI? Camerinelli suggests starting with a well-defined vision and specific problem identification. “Begin by examining data sources, ERP systems connectivity and data quality first,” he says. “Don’t view AI as a silver bullet – ensure you have proper resources, skills and strategy before deployment. Take a step-by-step approach with realistic expectations.”

Matsoukas, meanwhile, suggests that newcomers to AI “start narrow and data rich. Pick a single pain point, get the right data set and run a proof-of-concept with a trusted advisor and an established vendor.”

He adds that treasurers should also tap into their own talent for ideas and execution – “you may be surprised how many colleagues already experiment with AI in their personal time. Make the most of that creativity and energy.”