If only we could predict the future. Wouldn’t that be great? Life would also be a lot easier for treasurers and their liquidity management. The development of machine learning and artificial intelligence (AI) certainly seem to be taking things in that direction, but for now many challenges remain for treasurers hoping to take advantage of the latest technology.
In a recent webinar, Jean-Baptiste Gaudemet, Senior Vice President, Data and Analytics at Kyriba noted how data and analytics solutions have evolved over the last decade. Not that long ago, it was only reporting solutions with descriptive analytics that were available. And then came business intelligence solutions with diagnostic analytics. More recently there has been huge progress and development with machine learning with its predictive analytics. And then there is AI, which not only uses data to make predictions, but also uses prescriptive analytics to offer the best action that should be taken.
Despite these developments, however, treasurers aren’t close to being able to predict the future. Gaudemet says that so many treasuries are still relying on Excel. “Despite the development and the maturity of this technology we still see a lot of treasurers stuck with Excel reporting and they face a lot of issues,” he says.
The problems are well known; individuals are manually entering data, which is prone to error. The spreadsheets can also be highly dependent on certain individuals, which creates a problem when they leave the company – and take that knowledge with them. And, Gaudemet says, “On a spreadsheet you cannot implement sophisticated machine learning and optimisation algorithms. In the context of treasury that means spreadsheets basically give inaccurate cash forecasts which has a very negative impact on the business, especially nowadays with the current market volatility.”
There are many benefits and use cases for AI, such as optimisation tools that advise on the best action regarding cash movement. Although many treasurers agree that accurate cash forecasting and applying AI to manage liquidity effectively are desirable, there are many challenges that remain.
Gaudemet notes that the major challenges are a lack of data scientists, high expectations and budget. Companies face difficulty recruiting data scientists and they are in high demand. “It is very difficult to hire them and keep them in the company,” says Gaudemet. And this has a knock-on effect for the cost of a project. “As a result of this situation the budget will be very difficult to justify – that’s why we believe that the only solution is a packaged solution.”
This is in line with a Capco whitepaper on AI and data analytics, which also notes the lack of competence available for companies hoping to build their own solutions. The Capco research notes that a pilot project typically has seven people and it can be challenging for a company with multiple AI projects to find the people they need. Similar to Gaudemet’s view, Capco also notes that companies need to consider if it is worthwhile building their own solution and compare that to the cost of buying something that is ready-made.
The consultancy Capco also notes other challenges with the adoption of machine learning and AI. For example, AI cannot always account for unpredictable events as some things cannot be predicted by machines, such as unplanned outflows or inflows. Also, there is the machine algorithm bias to consider, notes Capco. A machine may be trained on a dataset that is limited and may learn to ignore certain events, which actually may have an impact on its output.
These are just some of the challenges that remain for treasurers who have moved away from Excel and are seeking the holy grail of accurate cash flow forecasting. For now – given these hurdles – accurately predicting the future still seems a long way off.