The transparency challenge: providing algorithmic clarity in the financial sector

  • BLOG POST
  • 07/02/2019
This article was originally featured in Finance Derivative.
Author James Loft, Chief Operating Officer, Rainbird.

As an industry, financial services is amongst the most data-heavy. The information flowing through firms is sensitive, diverse and complex, and covers everything from consumer credit card transactions to mortgage lending and stock price projections. Given its rapid rate of growth, big data has become an integral part of industry strategies to generate new revenue streams, enhance efficiency and improve the customer experience.

The result of this has been increased interest and investment in AI technologies. While it will come as little surprise that firms want to get more from their data, they’re also facing tough competition from new challenger banks and fintech companies, many of which are driving change faster than the more traditional players would like.

Previously, the sector’s primary use of AI tools has been to transform back-office processes. It’s been a cost-saver rather than a revenue generator, with applications such as robotic process automation (RPA) used to complete simple, repetitive tasks and internal deployment allowing for a conservative approach. But as the market evolves, a second wave of investment is pushing AI closer to front-line services, bringing with it some unique challenges.

 

Protection and regulation

Earlier this year, UK Finance and Parker Fitzgerald co-authored a report exploring the power of disruptive technologies and the sustainability challenges facing financial services firms in the digital age.

The report takes an in-depth look at the impact of new technologies on the sector, discussing how organisations should handle integration. This is of particular importance to complex financial organisations such as banks, as they are comprised of intricate infrastructures involving complicated risk and compliance processes. All of which are subject to intense scrutiny and regulation.

When it comes to AI, one of its biggest benefits is the potential to help firms better match internal policies and procedures with the external demands of regulators – for example, improving the way in which risk within the business is modelled and managed. New data-driven platforms can help firms become more sophisticated in the way they tackle issues such as fraud, identity theft and money laundering, allowing them to provide greater protection to the consumer.

 

Importance of auditable AI

In an industry as heavily regulated as financial services, auditable AI is of significant importance to firms. While organisations have been applying AI to business processes, most concerning to regulators is a loss of market transparency.

At present, the delivery of services such as mortgage applications and credit scoring though automated advisory models are some way off. This is because deep learning and neural network systems are not yet fully auditable. Their use creates a problem referred to within the technology community as “black box” decision-making.

But in regard to important issues relating to consumer protection, such as fraud, firms can deploy auditable automation technologies. These platforms use algorithms to spot suspicious activity, detect patterns and predict outcomes in large data pools. Some of the more advanced solutions are even capable of assessing the anatomy of a fraudulent transaction — they can draw inferences based on the information available, raise questions where the data is incomplete and produce audit trails.

What consumers and regulators want is a clear-cut way to interpret AI tools; to ensure data is being used both legally and ethically. So, for financial organisations applying AI-powered solutions, automated technologies that are auditable and explainable when challenged represent a huge step forward when it comes to compliance.

To underscore the importance of transparency within the sector, PricewaterhouseCoopers (PWC) has developed a responsible AI framework which outlines best practice principles and provides guidelines for firms to follow.

 

What’s next for AI in financial services?

While AI technology is certainly making its presence felt within the sector, adoption rates are still slow. The general feeling is that the industry is feeling its way forward – but change is coming. According to McKinsey, in the next few years AI will complete between 10 and 25 percent of the work across bank functions. Much of this is likely to continue to be in back-office processes, with a more measured increase in consumer-facing products.

So, what we’re likely to see as AI continues to reshape the financial services sector is a strategic approach to implementation. A streamlining of internal operations that helps firms match internal processes with external regulation. And the best way for firms to do this is to invest in auditable, transparent and explainable AI-powered solutions.