How AI can add value in treasury

Artificial intelligence (AI) has moved, in a very short space of time, from the latest industry buzzword to a practical, transformative tool that can reshape how organisations operate.

Even treasury teams, with their extensive reliance on sensitive data, high-levels of risk governance and high compliance thresholds, can benefit from AI.

From predictive cash flow management to real-time fraud detection, AI offers opportunities to enhance decision-making, reduce manual workloads and uncover insights that were previously out of reach.

With this opportunity in mind, HSBC hosted a session at the recent Association of Corporate Treasurers 2025 Annual Conference – bringing together a host of AI and treasury experts to explore existing practical applications of AI in treasury, as well as tips on how AI tooling should be deployed, monitored, and controlled.

Here are the key takeaways from the session:

Technologies that can support operations and processes have been evolving for some time – but there remains much confusion around the different iterations. Here’s a simple guide:

• Artificial Intelligence is the broad concept of creating machines that can mimic human intelligence – a kind of catch-all term for these technologies.
• Machine learning tools are a subset of AI that have been in use within businesses for many years. They enable systems to ‘learn’ from data without explicit programming. Electronic maps or travel booking apps are good examples – using previous journeys to predict where you might want to go in future.
• Generative AI (GenAI) is the catch-all term for AI tools that can create new content and ideas – such as text, images and projections – by mining and identifying patterns from vast datasets, to then generate a view based on those patterns.
• Large Language Models (LLMs) are the tools that Gen AI systems use to understand and generate human language. They are the tools that are trained to recognise patterns, understand context, and produce coherent and relevant responses.

Gen AI presents an important opportunity for treasury teams to improve their processes, reduce manual tasks, free up resources, and better interpret data. That’s because treasurers can ask it any question – even if it hasn’t been asked before – and receive a response that Gen AI has generated by accessing LLMs and large data sets.

While there are many ways in which Gen AI can support and enhance the treasury function, three areas are gaining the most attention at present. According to a recent survey by Deloitte, treasurers identified the most popular use cases within treasury to be cash flow forecasting, cash positioning, and FX and interest rate management.¹ This is because these areas rely heavily on analysing large data sets, combined with other available information, to make assessments or projections – tasks that Gen AI can carry out by leaning on LLMs.

Gen AI can also free up valuable resource across treasury – empowering staff to focus on more valuable and strategic tasks.

The session heard from one panellist who explained that using AI to analyse data and identify trends within spreadsheets has had multiple resource benefits: “AI can analyse those data sheets much faster than me, freeing me up to do other things. It can do it in a ‘no-code’ way, meaning I don’t have to be a technical programmer to use it. It can analyse the data far deeper and far better than me – because it spots trends and patterns that I might not. And it gives me both the time and the information needed to answer questions form cross-function stakeholders. That last point is really important if we want to become more strategic players within the organisation.”

It’s important to ensure all the right stakeholders are involved in the implementation of AI – while involving them in the right order could save wasted effort. Speak to the finance team first to identify the hardest pain points and identify what you want to achieve. Once decided, engage your compliance and the risk teams early – as they will ultimately make the call on whether an AI tool is appropriate for the task. IT teams are often the guardrails for security when it comes to data – and can quickly put the appropriate protections in place. So, engage them early, too.

While Gen AI tools will get better with time – ‘learning’ as they go – it’s important to ensure human checks are in place where appropriate. Take opening a bank account, for example. A Gen AI tool could, in theory, understand the account opening forms and, drawing on data and information you hold, complete the forms for you. However, you wouldn’t risk submitting them to the bank without a human checking them first. So, work out where in your workflow human checks and course correction should take place – and ensure it happens every time.

As Gen AI tools learn as they go, getting better with time, it’s important to manage expectations about what they will deliver, especially in the early days. One panellist told the session: “I often think of Gen AI like having a new analyst in the team. If you asked that analyst to compile a forecast on day one – when they have not done it in your organisation before and they have very little knowledge of the business and its strategy – you would expect that forecast to have gaps in it and you would check it thoroughly before forwarding it on. However, two years down the line, you would expect that analyst’s forecast to be much more informed and complete – and would have more confidence it. Gen AI tools are very similar in that respect.”

For treasurers not yet using Gen AI, a sensible way to begin exploring its value and what it can do for the treasury department is by experimenting with AI and LLM tools on a personal level. Experimenting with ChatGPT or other LLMs is a good way to understand how their prompts work, without putting your organisation’s data and information at risk.