< Back to Blog

Artificial Intelligence in Financial Services

Ness recently hosted a thought-provoking discussion in Manhattan for a selection of high-powered guests from within the Financial Services industry. The subject for discussion was the prevalence and maturity of those two, now ubiquitous letters: A and I. I was compere for the event and, as I looked out through the audience (and having seen the guest list), I saw representations from investment banks, wealth managers and hedge funds; and job titles from client relationship and trading leaders to data architects and Head of Analytics. We really had a cornucopia of interests, perspectives and appetite for innovation.

As the compere for the event, I set the stage by putting the Ness Point of View forward. In this space, through our work with many Financial Services companies and other adjacent (or even unrelated) industries, we see some fundamental truths which need addressing. In a world of AI-hype, there is enormous potential for misdirected efforts which lead on to unsatisfactory results, and these risks will only be mitigated by tackling these truths from the beginning.  Core to it all is structured data, talent and business-focused use cases. Misery awaits those who hire five data scientists and set them in a corner waiting expectantly for Sesame to Open. There are universal business principles which apply here as with every other hot topic: you need senior, open-minded C-suite support. You need ordered data with structure, hierarchy, taxonomy and ontology before you can start training an AI engine to look for patterns. And, you need top talent in leadership, management, execution and validation roles. This is a team effort, and cross-functional, complementary skills are essential for any AI initiative to take off and deliver commercial value.

The drivers for AI’s general growth spurt are frequently cited: low cost of enormous computing power and storage in the cloud, accessibility of open source platforms with commodity algorithms to personalize with your data, and the ubiquity of data generating sources like mobile phones and IoT sensors. To those I added some Financial Services-specific factors to channel the discussion:

  • Privacy, security and compliance with the regulator have been used historically as a reason not to move as fast as some of the Silicon Valley behemoths in AI (think Google Search, Skype Translate and Amazon/Spotify recommendation engines).
  • Regulations of the last 10 years mandate a move to “Digitize Everything.” Although this was intended to create audit trails for compliance purposes, it has been the catalyst for forcing structure on previously unstructured data sets: an essential ingredient for AI.
  • An appetite for finding “insight in usage” has meant intelligence is being uncovered from clickstream analysis of on-screen behaviour by consumers, analysts and buyers and sellers. This is now being augmented (and enriched) by data from voice calls and social media sentiment analysis. (Indeed, some AI-based investment strategies are based on volumes of retweets of a particular news item and now an interpretation of what the text in those tweets really suggests).

My guests on our distinguished panel included Shekar Pannala, CTO of S&P Global Ratings, representing the large enterprises and S&P Global, which generate value through managing and finding insights in billions of data points every day. S&P Global Ratings is focused on delivering what they are calling Essential Intelligence to “provide users with the tools to make critical decisions with conviction.”

Shekar was joined by Sean McDermott, Senior Analyst from Corporate Insight, an analyst firm that provides recommendations and expert analysis on improving the digital offerings and overall user experience to over 100 financial institutions. Sean writes widely on FinTech trends and how disruption in Wealth Management is moving into disruption in Insurance and Retirement Planning. He specialises in the Robo-Advisory space.

Last, but by no means least, I called upon Rajeev Sharma, Chief Solutions Officer at Ness, and one of Ness’s most passionate advocates for advancement and maturity in architecting these Platforms of Intelligence. Rajeev provided a technology perspective in his inimitable style and brings decades of relevant enterprise experience, coupled with his academic odyssey at the MIT Sloan School of Management and School of Engineering Systems.

I asked the panel about the landscape and appetite out there in the market for AI-inspired strategic offerings – and what were their observations on common obstacles holding companies back.

Shekar noted a home truth: that S&P Global Ratings works in an extremely regulated environment and that whatever AI solutions they create need to respect and not violate that framework. It may well be okay for certain Hedge Funds to openly attract investors to follow their bespoke AI-investment strategy, but S&P Global Ratings is used as a prime resource by companies, governments and individuals to make serious decisions about other people’s money, so their innovations must be ready for scrutiny from the regulators.

Sean added that the Robo-advisory industry was improving its offers rapidly. The likes of Betterment and WealthFront had set the ball rolling but had some of their thunder stolen when the giants like Vanguard and Charles Schwab responded with similar all-digital offerings. The battle is now on to offer more features that deliver on the need for “extreme personalization,” whereby the intelligence is taking account of known needs, behaviors and preferences, and advising, guiding and maybe executing on some of those personal investment drivers. Bank of America is pushing Erica and there is now Finn by Chase. The move to a personalized intelligent assistant advising you what to do next is accelerating fast. And, it will be part of the working day of every financial analyst very soon.

Shekar touched on the need for building solid business cases and technology accelerators as proof points ahead of making major investment decisions. He also stressed the potential from accessing and integrating seamlessly with new data sources and adjacent datasets. Indeed S&P Global Ratings is making investments in companies which specialize in satellite imagery which will surface insight and generate data for analysts in the commodities, transport and logistics markets. S&P Global Ratings is also making investments in automating as much of the ratings and publishing business as makes sense, removing inefficiencies and redundancies to help get accurate information out in near real-time.

Next, I asked about the nature and nurture of talent. Do you train it or buy it in or buy a company? There is a lot of jostling for position going on out there, and it seems that for many FinTech startups, being acquired by one of the big boys with all their scale and balance sheet is a more likely outcome than overtaking them in the market. S&P Global Ratings and Corporate Insights both see a large appetite for investing in talent and partnerships to make it happen faster. Rajeev made a strong point to sum it up: Ness has invested in training over 100 engineers to take them from being a strong engineer to be a full-stack, industry-ready AI engineer. The reason for doing that is because talent in this area is so hard to find with any scale. Many firms will struggle to mobilize a decent-sized team and get to market quickly with a solution for this reason. If they choose to do it internally, they may end up having to buy a company to get the talent – and they are expensive hen’s teeth at the moment.

I got to ask the big questions like were the days of the human analyst numbered, and whether a machine was likely to offer better investment advice than a human any time soon. Both Shekar and Sean smiled knowingly, and seemed to follow the same line. It may well happen – but not in the immediate future.

Driving the use cases that I heard about from the audience in the Q&A were demands from investors, customers and employees for extreme personalization. Based on intelligence latent in related adjacent data sets (but previously untapped and locked out of sight) is a market-driven need for real-time notifications, guidance, pattern spotting and calls to next best action to make my day (whoever I am) more fulfilling and valuable – and simultaneously, less mundane and less productive.

It was a stimulating discussion with a positive energy directed to making things better rather than noodling over a dystopian future where we have no say in the narrative. I very much look forward to continuing those discussions in subsequent conversations with new-found colleagues.