AI seems to be on the agenda for most companies and their treasury teams, and in response, the ACT held a webinar in May to investigate some of the practical experiences of getting started with AI. The webinar had almost 800 registrants, a testament to the interest in this topic by the treasury community.
To provide real-world insights, the panel brought together Tom Thorn, senior treasury manager at Personio, a treasurer who has been implementing AI for more than a year, and Gurjit Pannu, CEO and co-founder of Palm, a fintech company that has been automating treasury operations for several years.
We ran a poll during the webinar and found that only 10% of attendees either had a clear strategy or were already successfully using AI, with almost 50% identifying some use cases, and 28% still not clear where to start.
The survey suggests that while there is much discussion around AI, turning it into meaningful actions still remains elusive to many treasurers.
Here are the key takeaways from the webinar:
Tom set out with some clear objectives based on areas of pain for him and the C-suite. One project Tom undertook was to bring together a disparate set of data points that would finally give him visibility over his cash balances.
Gurjit stressed the importance of setting realistic deadlines. It is tempting to over-promise and many people make it look simple. Proper planning and setting realistic timeframes are crucial. Treasury teams are already busy so it’s even more important to be honest about how long a project could take, especially if there is a large amount of learning required.
Neither Tom nor Gurjit confessed to having perfect data but it hadn’t stopped them from moving forward on their projects.
Gurjit referenced minimum viable data as a useful starting point when deciding on the importance of data quality. He also shared the concept of bucketing data into varying degrees of quality and of using validation tools to continually improve poor data. This reflects the reality of most companies and some uses of the new tools available.
Tom acknowledged that it had been hard to know where to start. Listening to his story, it seemed that a good place was to consider some of the manual tasks that require data from unconnected systems and cause the most pain.
As we often hear, every treasury team is different and so it makes sense that the most valuable use cases may differ from one organisation to another.
Gurjit spoke about the value of using AI tools to improve existing processes rather than aiming for a big bang transformational approach.
A key concern for all AI users is being able to trace what outcomes are based on.
Both Tom and Gurjit were mindful of the problems with black boxes but both agreed that there was no reason to accept this. Outcomes could be provided through step-by-step decisions or through a single consolidated decision. The choice of which depend on various factors including the risks associated with the outcomes and company culture.
Gurjit referred to embedded explainability as a way of building in the ability to interrogate the decision-making processes. In some respects, it is no different to how treasurers have used TMSs – they may not be able to replicate the answers exactly but they understand how the answers are calculated and can provide estimated solutions.
Tom said that treasurers need to learn how to ask the right questions and in the right way. He shared a couple of examples of how he had used tools like Palm to speed up certain processes and to provide better visibility over others.
It was clear that, although Tom had started his journey with the typical technology skills of a treasurer, by starting with small projects that were easy to experiment with and fine tune, he was able to build up his confidence to do more projects and with more complexity. This brought him a deeper understanding of the capabilities these new tools can offer.
Naresh Aggarwal is associate policy and technical director, Association of Corporate Treasurers
Watch the webinar on demand by registering here.
