Even with machine learning, human judgment still required in fintech sector
A COMBINATION of recent events has seen a rapid acceleration in the adoption and incorporation of technologies by a wide range of firms and institutions in the global financial sector.
Whether this adoption has been spurred on by the global financial crisis of 2008; the need to adhere to regulation; or the immediate need to pivot and handle the consequences of Covid-19 and its impact on customers and staff, firms in the finance industry are embracing financial technologies (fintech) into their daily processes.
Designed to drive enhancement in services and improve efficiencies in back-office operations, it has seen a thriving sector developed beyond traditional ‘Wall Street' financing.
The prospect of the part that machine learning (ML) could play is generating a lot of momentum.
The financial sector is well-placed to benefit from machine learning, with large volumes of historical structured and unstructured data to learn from. It is also open to implementing new technologies, as demonstrated by the early adoption of technologies such as algorithmic trading by investment banks in the 1980s.
Accordingly, a study by Forrester in 2019 estimates around half of financial services and insurance firms globally already use ML technologies. By using these technologies, significant and non-trivial savings have already been made. For instance, JPMorgan Chase has estimated their fraud detection solution, which uses machine learning to analyse stock market data, saves the bank $150m annually.
So, will machine learning completely automate human tasks in the finance sector? Probably not. Human judgment is still required to help with so-called ‘edge cases', where no obvious outcome is clear, and associated decision-making.
In many ways, it represents a new synergy between human and machine. Machine learning systems can sift through enormous amounts of data and identify correlations. Human expertise is still required to tease out spurious links and noise from underlying informative signals. As highlighted by the Covid-19 pandemic, machine learning is highly capable in analysing large domain-specific data and identifying patterns to an expressed objective, but is slower to adapt to these rare ‘black swan' events if they are not closely related to past trends.
On a positive note, using these tools alongside human judgment can improve the quality of data analysis for decision making and increase process efficiencies. Two such areas where machine learning is having an impact include fraud detection, and improvements in personalisation for customer service.
As we look ahead post-pandemic, we can expect to see the finance sector continuing to adopt machine learning technology to improve efficiencies and reduce costs across customer service, regulatory adherence, fraud detection and trading.
Machine learning strengthened with human expertise at this stage will aid in the development of more robust technology solutions.
:: Fiona Browne, head of AI at Datactics, will chair a panel discussion on artificial intelligence in financial services at AI Con this Thursday and Friday. To register for the conference go to: https://aicon2020.com/