Artificial intelligence has arrived in finance
Digitalization, Trends and Technology, Payments and Financing
Over the past few years, Robotic Process Automation (RPA) has begun to revolutionize knowledge work in finance and accounting. Today, software robots are performing many routine tasks that were once thought to be impossible to automate. But the real disruption is still ahead…
“By definition, a disruption destroys the old and creates something new. Now the technologies that utilize Artificial Intelligence are capable of doing this,” says OpusCapita Head of Product Marketing Petri Karjalainen. “So far, RPA has only helped to make current processes more effective. But now machine learning and predictive algorithms have started to create whole new possibilities for increasing automation in the finance and procurement processes of companies. And this is not just a vision for the future – but the reality of today.”
OpusCapita is already piloting machine learning algorithms with selected customers in the reconciliation and posting of purchase invoices. The solution will be made more widely available in early 2017.
Automation that overcomes the deviations
Despite the high level of automation in companies’ account payable processes, posting supplier invoices typically requires a lot of manual handling. Processes that include a lot of deviations and require complex rules and understanding of the context are generally not suitable for robotic process automation.
“A software robot needs clear rules set in advance, based on which it performs a certain task,” says Head of RPA & AI Jaakko Lehtinen from OpusCapita. “Self-learning machines can overcome this challenge with a machine learning algorithm that can create the rules and the logic as the basis of its operation by processing thousands, or even millions of rows of historic data. Basically, it learns to perform the task based on the decisions that humans have made previously.”
Instead of having to check and enter the cost center and other accounting dimensions for the invoice, the accountant will receive a suggestion for the probable correct posting. They can then either approve the proposal or make some corrections, in which case the algorithm will adjust through the feedback. The next time, it will be able to suggest an even better solution.
Success rate of 95 percent
In addition to purchase invoice posting and reconciliation, intelligent automation with AI can be applied to many other parts of the extended purchase-to-pay chain. According to OpusCapita’s Data Scientist Ali Faisal, machine learning algorithms can be utilized to detect deviations and anomalies, for instance, as well as to classify and categorize things, or to generate predictions.
“Preventing payment fraud is a topical challenge in organizations,” says Faisal. “With machine learning, we can spot the payments that deviate from the norm in the payment flow and have them checked for fraud or for mistakes before they are actually paid. Predictive algorithms can use data from various systems to generate forecasts for cash flow, sales or even demand.”
Faisal emphasizes the importance of acquiring the right kind of data for the machine to learn from and base the rules on. With machine learning, automation improves continuously through trial and error.
“Based on our experiences, the machine learning algorithm is able to suggest the correct posting dimensions 95 % of the time,” he says. “Interestingly, financial departments are finding it difficult to accept this probability, although it is most likely a clear improvement on the performance of personnel! Historical data often reveals that similar supplier invoices have been posted in several different ways by people.”
Day-to-day machine learning
Intelligent automation utilizing machine learning and AI will advance at a far faster rate than people are realizing
"Previously, the lack of cost-efficient machine power and the limitations of the server environments, for example, were hindering the wide-scale adoption of these solutions,” says Jaakko Lehtinen. “Now the situation has changed, and development is starting to pick up speed.”
OpusCapita’s experts are anticipating that machine learning will soon be a part of the financial professional’s day-to-day work.
“People in finance are particularly keen on using spreadsheets and macros in their everyday work, for instance. I believe that in no time at all, these tools will be replaced with machine learning algorithms,” says Petri Karjalainen.
Machine learning and artificial intelligence were topics at the Finance 4.0 event organized by OpusCapita and Talouselämä in Helsinki in November.