Artificial intelligence has a decades long-history with the financial services industry – the two fields share not only a symbiosis in pushing forward domestic and international progress, but also, in recent years, has shown how the potential to disproportionately distribute the spoils of innovation. Intertwined with the process of digitisation, the use of AI in financial services is anathema to many in the United Kingdom.
The recently commissioned FCA report produced by the Alan Turing Institute aimed to ‘inform and advance the debate about responsible AI in the context of financial services’. In this piece, I aim to briefly introduce the three distinct elements of AI innovation outlined in the paper, namely automation, non-traditional data, and finally, machine learning (ML) overlaying contemporary examples and immediate challenges at hand beyond the initial discussion in the report.
Unlike the other two pillars of innovative AI, automation is not a particularly new phenomenon in financial services. Examples of symbolic AI are well-known, such as the process used to identify unusual transactions, a mainstay of back office operations in preventing fraud and crime for decades. It is an example of implementing an AI technique, whereby the customer receives increased protection by virtue of additional security exercised by an initial digital flag, and subsequent human intervention.
Automation fundamentally reduces the role of humans in the performance of tasks and processes. For many retail customers, where the predominant relationship with financial institutions is in the banking and lending space, it has long been the case that the financial services industry has doubled down on the finance but has long left behind the ‘service’ element of financial services. This stands to become even more acute with the predicted demise of the branch-based banking model, with 65% of 305 senior global executives surveyed .
But do we need the branch? The pandemic has proven, arguably, no. In the UK, the Bounce Back Loan self-certification process changed that perspective for many of the digitally cynical in the United Kingdom, with hundreds of thousands of businesses received access to capital en masse from banks through automated decision-making, information management and information verification workflow processes – a scenario unthinkable in traditional analogue relationship management.
However, we must ensure that automation does not compound digital exclusion. The report notes ‘adopting AI solutions can be accompanied by an increased reliance on Digital communication and service provision’ and can contribute to a sense of disempowerment for certain individuals. More must be done to bring all of society along on the journey across generations and social divides; a crisis has proven that adoption is possible, but it can only just be the start.
The quality of data is fundamental in the application of artificial intelligence. Where traditional measures fall short, firms have been looking for alternative data points to inform their processes and most commonly, their decision making.
Out of the three innovation pillars it is arguably the most ideologically contestable – from the perspective of data protection, public transparency and the social impacts that have the potential to lead to unforeseen harm. The report notes that ‘the use of AI may raise concerns about individuals’ ability to make informed and autonomous choices and exercise meaningful control over their lives’. Indeed, the European Union set out the usage of alternative data points to inform social scoring and, by extrapolation, credit scoring decisions as unacceptable – by virtue of . This is in stark contrast to China where the use of alternative data points to nudge and inform decision making in financial services is most prescient in the payments space.
The seamless technological experience of the ‘all-in-one’ applications devised by Tencent and Alibaba utilises data points across the company’s corporate ecosystem and delivers a seamless user experience. As Gottfried Leibbrandt, former CEO of Swift has noted, the fact that the Chinese make four times as many digital payments than the West on average shows that they are ‘. In the United States, the tech giants’ inclination to replicate their own ‘one-size fits all’ applications appears ominously against this.
The jury remains out in the United Kingdom, but, with widespread concern about the utilisation of data through tracking and tracing over the past 18 months, the usage of non-traditional data will be limited in Europe and the United Kingdom precisely for the reason that, here, remains an innate and entrenched value system that prioritises the preservation of personal data and transparency in market processes. Whilst this will denude over time, through generational shifts, I predict the foundations of such a value system will stand in the way against data capture in the name of unfettered optimisation.
As defined by the report, machine learning (ML) refers to the development of AI systems that are able to perform tasks as a result of a ‘learning’ process that relies on data – contrasting with traditional approaches of embedding explicit rules and statements into code.
In an increasingly complex market, the use of machine learning has the potential to democratise financial services, particularly in the investment management space – reduced operational costs remove the previously high fee barrier to entry for increasingly sophisticated products – the success of eToro in both the US and the United Kingdom is testament to that assertion. In recent years, the charge has been that AI can act with increasing speed in the identification of new market patterns in extant data sets whilst removing the negative effects of human bias on investment decisions. However, at what cost to market stability, transparency and oversight?
An increasingly important consideration is that an ever-increasing complexity in market activities brings questions of comprehension not just for external stakeholders in understanding the processes involved, but more pertinently for internal stakeholders, whose responsibility is of an oversight and control function to the end client.
The report did not cover the regulatory risks; but it is clear that from an operational perspective, the risk and control frameworks of firms employing artificial intelligence, through increasingly complex mechanisms must ensure that there is a validation and verification process for activities which are underpinned by an equivalent level of human competency development in tandem at both the practitioner and regulatory level with improved AI practices to adequately perform oversight functions.
This should not just be restricted to the second line or mid to back office. In the case of an asset manager, maintaining one’s fiduciary duties as a portfolio manager throughout the client lifecycle is also required. In order to holistically undertake one’s activities as a manager, increased substantive, procedural and governance information will have to be incorporated into one’s investment decisions. In short, employing ML can exponentially improve the analysis of data sets; it requires a complementary level of human judgement to implement the findings appropriately.
For a nation which has poor financial literacy rates, what benefits are there to making an increasingly complex financial system more complicated? How, as a customer, will increased use of artificial intelligence and oft-voiced concerns of deteriorating relationship management practices be reconciled in a burgeoning decade of corporations assuming greater social responsibility for their actions. The proponents of artificial intelligence will cry it is the ‘Democratisation of Finance as we know it’, but how do we ensure socially beneficial innovation will occur?
This, for the financial services industry in the United Kingdom, and particularly for the ‘big five’ is the Polanyian double movement which it must address in the years ahead – managing the human element and ensuring that adoption and inclusion mirror with advancements and optimisation in Artificial Intelligence.
Written by Harry Wilby, SAMI Associate
The views expressed are those of the author(s) and not necessarily of SAMI Consulting.
Achieve more by understanding what the future may bring. We bring skills developed over thirty years of international and national projects to create actionable, transformative strategy. Futures, foresight and scenario planning to make robust decisions in uncertain times. Find out more at www.samiconsulting.co.uk.
Image by Tumisu from Pixabay