From prediction to action – how AI is changing treasury

The sheer quantity of data now available means it is too much for the human brain to analyse. At the same time, the need for such analysis to be done at high speed has never been greater.

Artificial intelligence (AI) can improve the ability to forecast future liquidity, but a more important development will be the ability of AI to support confident decision-making using the output from those predictions.

Artificial Intelligence (AI) has two rather different applications in the treasury world: machine learning to improve the prediction of future liquidity, and optimisation of the actions arising. There has been – quite rightly – a lot of focus on the former, but in our view, the latter is where the long-term benefits of AI for treasurers will really lie.

“Prediction” means the ability of a treasury team to forecast future cashflows over time with the help of statistical machine-learning algorithms. That timeframe can be short-term (less than one month) or long-term (typically one to three years ahead). Such prediction is important because successful forecasting enables treasurers to minimise their excess liquidity and then make the best use of that surplus in a variety of ways.

However, prediction beyond the short term is currently only possible for a tiny minority of companies. According to consultancy IDC, less than 5% of corporates can forecast cash reliably beyond three months, and less than 20% can forecast liquidity beyond one month. And, according to the Association of Chartered Certified Accountants, 90% of Excel spreadsheets contain errors.

These are significant problems for CFOs who are under pressure to improve forecast certainty but have reached the limits of manual cash forecasting and intuitive, human decision-making.

This comes at a time when the need for precise, accurate forecasts is at its highest level in recent memory. Inflation and interest rates continue to rise, increasing the cost of financing while also increasing the return on excess cash. Optimising liquidity, through accurate forecasting, is creating a major competitive advantage.

To improve forecast accuracy, finance leaders are relying on AI to leverage the exponential volume of data accessible to their teams. IDC forecasts that by 2025 the global data sphere will grow to 163 zettabytes (that is a trillion gigabytes). That is ten times the 16.1ZB of data generated in 2016. CFOs need a programmatic approach to harness their data and improve forecasting.

In a similar way, thanks to AI, treasurers are increasingly able to forecast their group’s liquidity at a particular moment in time based on the probability of the various cashflows throughout the business. Using output from the company’s treasury management system and enterprise planning systems, AI can analyse historic cashflows

Turning insight into action

But prediction is only half the story; the next step is taking action. Once a treasurer has confidence in the team’s forecasting results, they can decide how to invest any excess liquidity, whether in traditional money market funds, higher yield investments or in modernized programs such as dynamic discounting.

Similarly, knowing at exactly what points liquidity shortfalls are predicted to occur, the treasurer can also optimise borrowing decisions to minimize the cost of financing – or determine if simply negotiating new payment terms with suppliers will meet liquidity needs. These decisions have huge potential impact on the company’s results as well as its ability to manage risk efficiently.

In fact, our clients have demonstrated that they can save up to 50bp without compromising access to liquidity. That saving comes from increased returns on idle cash and investments alongside a reduction of fees and borrowing costs. In addition, the use of AI to solve problems like liquidity forecasting can free up time that can be spent on higher value-add and frankly more rewarding tasks!

Data analytics to enhance visibility, predictability and actionability

There is an increased demand for data analytics and data science capabilities in treasury platforms to help treasurers and CFOs optimise the management of their data while improving cash predictability and enhancing the actionability of the data.

For example, artificial intelligence algorithms that forecast short term cash flows and cash budgets automatically learn the cash flow seasonality and customer payment behaviour from the historical time series of cash flows (across all banks) and confirmed receivables (across all ERPs), enriched with public (e.g. bank holiday weather) or business (e.g. marketing campaign) data.

It is also possible to use AI to enable CFOs with the financial planning inputs to extrapolate the cash forecast to the long term, or to bring together all the cash, debt and risk information of a client thanks to business intelligence application. This would allow treasurers to analyse their net debt and liquidity available bank by bank or to stress test the forecast with various “what-if” scenarios. Finally, data analytic modules can improve the end-to-end treasury chain, from receivable to payable which can be optimised with a decision-support tool to allow treasurers to specify priority rules and ensure invoices are systematically paid on time and within budget.

Overall, the visibility and the actionability of the decisions made on cash will be improved on all timeframes from 7 days to 3 years thanks to these new capabilities. These end-to-end cash management modules are strategic assets that innovate the practice of cash and liquidity forecasting. Finally with AI it is possible for corporates to reduce their vulnerability to rising rates on a net basis, especially as they will continue to build cash reserves in anticipation of an economic downturn.

For more information: www.kyriba.com

[1] Kyriba Enterprise Liquidity Management Survey, IDC, July, 2021

https://www.kyriba.com/wp-content/files_live/content/kyriba-idc-wp-elm.pdf