Capturing Digital Transformation to Deliver Smarter Treasury in Ireland
By Ben Poole, Writer, TMI
Global digital transformation has kickstarted the Fourth Industrial Revolution, also known as Industry 4.0. For corporate treasurers, this represents a complete shift in approach to treasury processes, technology and, most importantly, talent.
HSBC’s recent Global Liquidity and Cash Management Forum, held in Dublin, Ireland, highlighted the technological advances that are revolutionising the financial services landscape around the world. For treasurers, these developments offer exciting opportunities to enhance efficiencies, but also present challenges in dealing with new technology providers, and securing businesses from cyber threats.
Alan Duffy, CEO of HSBC Ireland, gave the welcome address. He began by noting the excitement around technologies such as machine learning and artificial intelligence (AI), and how the financial sector is keen to find practical applications of such technologies. In an environment where interest rates are low and will likely remain so for some time, it is incumbent on financial professionals to manage cash more efficiently and to make sure that a return on capital can be achieved more efficiently, he said. Technology is a key enabler in these areas.
Duffy also noted that while neural networks have been around since the 1950s, the big game-changer today is the amount of data in circulation, as well as the availability of open source software and increased processing speeds. This enables data to be used and interpreted more effectively. Duffy said the largest spenders on AI and machine learning are banks, and the technology that they are investing in has to be deployed to enable treasurers to manage their function more effectively. He cited the launch of HSBC’s next generation virtual accounts as a practical example of this, reducing the multiplicity of physical bank accounts that corporates might have, making their jobs easier.
The era of Treasury 4.0
Following Duffy’s welcome address, Lance Kawaguchi, Managing Director, Global Head – Corporates for HSBC’s Global Liquidity and Cash Management business, took to the stage. He explained the context of how treasury is evolving as the world enters the fourth industrial revolution.
The good news for treasurers, he said, is that the new treasury reality, Treasury 4.0, is bringing several innovations to transform the landscape – and prompting new regulations, such as the General Data Protection Regulation (GDPR) and the updated Payment Services Directive (PSD2), which are revolutionising the way data is collected, stored and distributed. The instant payments landscape is also bringing further opportunities for innovations that corporate treasurers might wish to take advantage of, such as Request to Pay and Swift gpi.
Despite the digital opportunities, cybersecurity has become a risk management imperative. Companies need to continually invest in the latest technology protocols and firewalls to protect their devices, their networks and their reputations, said Kawaguchi. For treasurers, this means ensuring that companies’ cash balances are kept safe from cyber criminals.
Kawaguchi also warned that treasurers must look beyond the hype of Treasury 4.0 towards practical, real-world applications of disruptive technologies. He cited HSBC’s Liquidity Investment Solutions Portal, called LIS, as an example of this type of technology in action. It enables treasury functions to invest surplus cash automatically according to predefined parameters, which he noted was a powerful tool for the real world of Treasury 4.0.
The power of data
Peter Simon, Lead Data Scientist at DataRobot was next to address the event. DataRobot is a leader in automated machine learning, with a platform that enables data scientists to be exponentially more productive but also allows other users, those outside the traditional machine learning audience, to easily build and deploy sophisticated, robust machine learning models.
A typical AI system is made up of multiple components. Alongside databases and rules-based decision systems, the bulk of a typical AI set-up often consists of machine learning models.
There are two main types of machine learning. The first is ‘supervised learning’, which accounts for some 80% of machine learning applications in business. Supervised machine learning is about building models using historical data with known outcomes and using these models to predict what will happen with incoming records. In a given body of data, each row will represent a record, event or item, e.g. a transaction, a customer account, an invoice, or an order. Each record will have various fields (or columns) associated with it. For example, for a transaction, there might be a column for the payment amount, a column for the payee, some columns detailing information on the payee or the products sold, and so on. Alongside that sits the “target variable”: an outcome that the user wants to predict—was the payment fraudulent, did the payment arrive successfully, was there an operational failure in there, did I sell a product to a particular customer, for example?
The other type of machine learning is ‘unsupervised’. This is where the user doesn’t have an outcome to predict; rather the data itself is being examined for similarities and differences. Combinations of supervised and unsupervised machine learning can be very powerful (for instance, in fraud applications, where the supervised learning can detect known patterns of fraudulent behavior, augmented by unsupervised learning to detect anomalies which may be unfamiliar).
At the highest level, three things are required for machine learning:
- Things to predict — accounts, events, customers, transactions, cases and so on
- Data describing these things — do you have (or can you source) data which describes the things you want to predict?
- Business value — is the ability to predict likely outcomes of the incoming data, etc, valuable to the business; or, would having a good model of the behaviours described by the data be of strategic value? Some machine learning use cases are philosophically very interesting, but don’t produce anything of tangible value.