According to Nick Pedersen, head of digital at NatWest, corporate treasurers will increasingly benefit from AI’s growing accessibility, low cost of building models, and the sharp rise in investment, especially since the release of ChatGPT technology in November 2022.
Pedersen said AI can be used for data transformation, in which unstructured and disparate datasets can be set into structured form, and summarisation, where vast pools of content and data can be distilled into simple text, as well as transcribing speech to text and vice versa.
It can also be used for prediction, in which forecasting is based on historical and inferred data, and content generation, in which new images, text, code, and sound can be created from existing datasets, he added.
At a moment, I can see where our exposures are, including through joint ventures. It means I don’t always have to go to banks to ask what things I should be worried about
Pearson’s James Kelly revealed his AI journey "started four to five years ago”. Kelly, who is SVP treasury, risk management and insurance of the FTSE 100 education group, said “undertaking trillions of transactions in over 100 countries was making it harder and harder to keep an eye on what really matters”.
In response, Kelly and his team have developed a dashboard for real-time counterparty monitoring from a combination of generative AI and scraping by programming language Python. “At a moment, I can see where our exposures are, including through joint ventures. It means I don’t always have to go to banks to ask what things I should be worried about,” he said.
The value of AI was expressed as offering efficiency, prioritisation and the means to narrow down uncertainty, by Marc Jagger, Barclays’ director of corporate analytics. “Its ability to generate insights, with the right health warnings, can help you get to where you want to go,” he said.
However, a session at the recent ACT annual conference revealed that 72% of attendees said they were still using Excel spreadsheets for forecasting.
Royston da Costa, assistant treasurer at Ferguson, said although the plumbing and heating products distributor continued to use Excel in forecasting, he said change was inevitable: “We need to start looking at using AI to improve that process,” he said.
“A lot of corporates may not be in a situation where they are quite ready to implement a bespoke solution, but they will need AI to fix the bits that are very manual and for forecasting longer,” Da Costa added.
Although DIY retailer Kingfisher does not currently use AI on the web-based forecasting platform it implemented two years ago, the group’s assistant treasurer Muneeza Mansuri, said: “It’s definitely on our automation roadmap.”
Andreas Gefnider, director solutions consulting at cloud platform provider TIS, said that the right foundations need to be put in place to fully capture the value of the technology. “AI needs massive amounts of data to recognise patterns to be able to predict things more reliably,” he added.
A cautious and safety-first approach to AI tools is really necessary
NatWest’s Jon Horrocks, head of FX options quantitative analytics, and head of machine learning quantitative analytics Tom Underwood, who have been experimenting with a chatbot called Merlin, warned of the risks posed by AI.
Data loss from network attacks, manipulation through AI interfaces, bias from underlying datasets as well as fake or harmful content from bad actors, were among the risks cited. Models lacking in explainability and questions over ethics in practice, were also referenced.
“A cautious and safety-first approach to AI tools is really necessary,” said their colleague Pedersen.
Lawrie Holmes is a freelance business and finance journalist. All comments were made during presentations at the recent ACT annual conference.
The ACT Annual Conference 2025 will be held at ICC Wales 20-21 May. Register your interest