Artificial Intelligence will help CFOs take better decisions

Artificial Intelligence (AI) is finding its way into various sectors and domains. It is essential that CFOs look into the possibilities of this technology as well. Professor Kristof Stouthuysen explains why.

Self-learning systems

Kristof is expecting that machine-learning, in particular, will have a great impact on financial applications. Machine-learning is the ability of computers and algorithms to identify patterns in data and to learn from these autonomously, i.e. without additional programming, to improve the algorithm and its performance. “Machine-learning will help financial professionals to analyse data better, make more accurate predictions, and make decisions quicker and better. The human brain can only work with a limited number of parameters simultaneously, whereas the calculating capacity of modern computers is almost limitless.”

Incidentally, like big data, artificial intelligence and machine-learning are no new phenomena. “However”, he says, “the exponential increase in available data is a recent development, and machine-learning algorithms are useful to analyse those large volumes of data. Furthermore, algorithms are becoming increasingly sophisticated, and computers have ever more calculating power. These developments bolster each other, which means that computers can process more data so that the algorithms can improve as well.”

Value creation and risk management

The activities of CFOs and their teams can broadly be divided into three categories: (1) financial, fiscal, and accounting transactions that are ideal candidates for automation, (2) risk management, and (3) value creation. A CFO’s role is increasingly focused on strategic decision-making and financial support because this is where they can deliver the most added value. “Determining the projects, activities, and actions that are strategically interesting for an organisation, as well as creating value, are important decisions, which should take into account as much information and as many parameters as possible. This is where machine-learning can help”, explains Kristof.

Machine-learning algorithms are also essential tools for risk management, in particular fraud detection. “For financial institutions that want to detect credit card fraud, there is no question of analysing spreadsheets line by line”, Kristof smiles. “After all, millions of transactions are made every day, so it would be like looking for a needle in a haystack. Today, machine-learning, and in particular unsupervised machine-learning1, is incredibly good at detecting anomalies, such as unusual deposits in an account or transactions with the same card at the same time at different locations. The learning algorithms can find the needle, and then an employee can look into the flagged-up transactions to see if they are fraudulent.”

Focus more on artificial intelligence

If CFOs want to play their roles as business partners fully, they must take more initiative in terms of artificial intelligence, believes Kristof. “We are seeing that departments such as marketing, operations, and HRM are already making great use of AI. CFOs and their teams deal well with structured data in particular, such as those from ERP systems. Conversely, they do not yet use unstructured data such as tweets, e-mails, blogs, vlogs, and various other sources of information in the public domain, and yet these kinds of data are growing enormously. However, most of all: artificial intelligence and machine-learning are the ideal tools to analyse these kinds of data to benefit from the sort of information they contain.”

Misunderstanding

Why do so many CFOs seem to be lagging behind? “I don't want to generalise,” answers Kristof thoughtfully, “but CFOs are often overburdened. They have so many other things to worry about that there's no time and space to get started with artificial intelligence and machine-learning. Sometimes it can also be down to a somewhat conservative attitude often related to their age: the average CFO is a bit older and not always au fait with the latest digital technologies. Moreover, what also plays a role, in my opinion, is that so many small and larger companies have jumped on the AI bandwagon to try and sell their services. CFOs may fear that an AI project will turn out to be expensive as an external party has to be involved, on whom they then may become dependent. So, let me clear up this misunderstanding: many of the new tools are open source. That means that as an organisation, you can simply make a start with AI — all you need is to be open to it and have a sufficiently entrepreneurial mindset.”

The same language

Of course, every technology comes with risks and challenges, and you must not be blind to these. The GDPR is there for a reason and may make it difficult to use specific data. Furthermore, processing and storing large volumes of data has consequences for IT infrastructure. But, concludes Kristof resolutely: “While you need to think about these issues, they are no reason for not using artificial intelligence”.

There is another misunderstanding that Kristof wants to get rid of: “Belgian CFOs often ask if they need to retrain as data scientists. They then tell me that they already have a team working on business intelligence. Of course, they do not need to retrain. CFOs and their teams have very specific knowledge and expertise, which they must cherish, but they also need to be able to work closely together with data scientists, and then you need to speak the same language.”

Get started

Of course, there are CFOs and financial professionals that already want to, or are, finding out more about AI, and that is good news: young people are more and more interested in technology such as artificial intelligence. In the war for talent, if a finance department wants to stay attractive, it must invest in AI alongside traditional accounting and finance skills.

In short, CFOs must take the plunge with AI. “Start with a small project”, advises Kristof. “The learning curve with AI projects is incredibly steep, but it is worth the effort. That's also what we see at Vlerick. For example, in Accounting & Finance, my colleagues and I are increasingly focusing on new technologies and how these will help financial operations to move forward.”

1 With supervised machine-learning, the algorithm learns on the basis of input-output pairs. With unsupervised machine-learning, the algorithm must find structure in the input. Unlike with supervised machine-learning, the output desired from the input is not provided during the training process.

Related news

  1. The advantages of virtual learning environments for Business Schools

    Date: 19/06/2019
    Category: Opinions
    Distance learning is not a new concept. But as time goes on, technologies continue to evolve and broaden capacity, ever improving the standard of online learning. The best online MBA programmes offered today are arguably just as good as an on-campus MBA, and can prove a better fit for a number of students. In today’s fast-paced working world, online MBAs offer flexibility and diversity that benefit both Schools and students, says Professor Steve Muylle.
  2. Digital transformation in Flanders: We should aspire to be number one, not just one of the top three

    Date: 28/05/2019
    Category: Opinions
    Digital transformation is a means of survival in the turbulent times that we are currently experiencing and that still lie ahead of us, according to Professor Stijn Viaene. “We have to make this cumbersome tanker change course. We may need to transform it into a fleet of sailing boats - provided that we give everyone the chance to stay on board. Because not only is this our damned duty; it is also the only way that we will be able to pull through.”
All articles