Machine learning in Financial Services. What are the opportunities?

The Bank of England and the FCA recently jointly published a report into the adoption of Machine Learning in Financial Services. In it they found that the new technology is quickly transitioning from emergent to mainstream. The study cites that two-thirds of financial services firms surveyed already use at least one ML based solution in some form, with the median usage being two applications. It is expected to more than double in the next three years. Mark Carney, Governor of the BoE, said in his speech at this year’s Lord Mayor’s Banquet that a new economy is emerging, to be served by a new finance, one that needs to balance transition with resilience.

Defined by the MIT as a category of Artificial Intelligence algorithms that employ statistics to find patterns in massive amounts of data, machine learning has permeated all aspects of day-to-day business and indeed the speed of development is staggering. One area causing excitement at the likes of Google’s development labs is Natural Language Processing. The technology behind the search engine’s often seemingly intuitive responses to search queries reached new milestones, with the recent integration of Google Bert , advancements in their effectivess to detect and correctly process the subtleties and nuances of language. For example, identifying the context around keywords to decode any modification, amplification or negation in an entry’s overall meaning.

Machine Learning applications for Financial Services

In the financial sector, the applications of ML have certainly ballooned in recent years and has untold potential to unlock the value held in large stores of data. There are numerous fully functional examples up and down the value chain, from back and mid-office operations through to customer-facing solutions either deployed to support human agents or as a direct customer interface.

JP Morgan’s investment in fintech Limeglass is a recent example utilising ML and natural language processing, in this case to facilitate institutional research, trawling documents for volumes of unstructured data to develop taxonomies and potentially slashing manual hours spent on research.

HSBC is employing the technology in the war on financial crime, with its Cog-I system screening hundreds of millions of transactions, doubling the speed and accuracy of sanctions and money laundering screening in cross-border transfers.

In insurance, software house OpenGI has announced a partnership to enhance their proposition through machine learning and predictive analytics. While this is a nascent venture, some of the more developed ML use cases in insurance include the much-documented Lemonade, and Zendrive, which combines machine learning, telematics and analytics to measure driver safety and traffic conditions.

Further along the value chain, and closer to the customer interaction, AI and machine learning continue to power increasingly conversational chatbots, such as those used by MasterCard among others, designed to minimise the friction of digital user interface in financial services.

A recent study by PwC highlights the growth in assets managed by robo-advisors, with Vanguard topping the list at US$115b AUM in robo-advisory. This is only set to grow – according to Anthony Cowell, Head of Asset Management, KPMG, “The rise of machine learning will really make our industry unrecognizable in the future.”

Human and Machine – the hybrid approach

Machine learning technology lends itself particularly well to detecting patterns in large volumes of unstructured data and has transformed the way organisations monitor and measure their reputation. It powers alva’s Reputation Intelligence technology, allowing faster processing of the unending deluge of media and other unstructured data that is being generated daily. As many adopters of the technology have established, this vastly enhances processing efficiency, but is not always an adequate stand-alone solution. A human layer of analysts – in alva’s case, sector expert teams applying years of industry experience – interrogates, validates, builds on and interprets the initial analysis, completing a hybrid approach, which cannot be substituted by a purely digital method. As Philip Watson, Chief Innovation Officer at Citi, says, “It’s a human plus machine world. It’s not a machine-only model. Nor do I see it becoming a machine-only model for a long, long time.”

While there are those who have highlighted concerns with the potential of emergent technology to displace human expertise and dehumanise services, machine learning has entered the mainstream. It is set to transform the financial sector by changing consumers’ expectations of how these institutions interact with them and provide them with valuable products and advice. Equally it has opened up new opportunities for organisations to leverage and extract value from previously untapped pools of data, both held internally and in the public domain.

So far, some of the most compelling use cases are those where organisations employ a hybrid approach, harnessing the efficiencies offered by machine processing and allowing their people to then apply the interpretation, critical thinking and judgment artificial intelligence cannot sufficiently replicate – at least not yet.