Panel discussion hosted by Fintech Influencers - comprised of leaders from across FS invited to come together by The Realization Group and The Harrington Starr Group - virtually, on Wednesday 22nd April, 2020. Led by Clive Posselt of The Realization Group, and featuring Matthew Squire (Founder of Fuzzy Labs), Josh Rix (Director of Woodhurst), and Angelique Dwyer (Principal Consultant at The Realization Group).
This was our first Fintech Influencer’s Online SIG, and we have been very pleased to have received such overwhelmingly positive and enthusiastic feedback! As a result, we will be organising a longer follow-up event – watch this space for dates and more information. If you missed the panel discussion, you can watch the video at the bottom of this article. In the meanwhile, here’s a summary of our discussion, to whet your appetites…
We began by framing the key questions for the panel to address:
1. How advanced is the use of artificial intelligence and machine learning (AI / ML) in financial markets?
2. Is the real use of AI actually less than firms would like us to believe – is much of it just down to smoke and mirrors and clever marketing?
3. What is AI actually being used for within financial markets operations?
4. What do organisations need to do in order to prepare for more widespread adoption of AI, and to set themselves up for success?
When we consider the potential for AI, it’s important to recognise the myriad forms that it can take, as well as their applications. We can start with optical character recognition – taking paper documents, scanning them and translating them to a digital format. Natural language processing (NLP) techniques can then be applied to these digitalised documents. This leads on to sentiment analysis, scanning documents for positive vs. negative messages and sentiments which can be applied as inputs to automated recommendations for investment decision-making.
Within financial services, these AI capabilities can then be applied to specific use cases. NLP is being actively used to implement chatbots in retail and investment bank operations, and generating pre-emptive responses to augment sales and customer support chats. AI technology has also been used to automate reconciliations and trade processing, leading to a reduction in exception management and remediation costs and overheads. UBS is using AI to detect payments fraud by automating traditionally manual AML and KYC processes (with their associated high FTE requirements). BNP Paribas led the pack by a few years with its SmartChaser trade matching tool, built using AI and predictive analysis. The tool predicts the likelihood of a trade not matching and requiring intervention, and even generates a suggested email to counterparty. Other work in banks such as HSBC has focussed on client email routing and generation of automated responses.
Sadly, for those holding the purse-strings on AI projects, cost savings and efficiency take priority, in many organisations, over the client experience (with regulatory drivers always coming out on top). This reflects the wider challenges in articulating ROI as it pertains to client service and satisfaction, despite this being an area that can benefit enormously from use of AI / ML to augment sales and operations interactions with clients.
There’s an interesting distinction to be made between trading technology and more customer-facing solutions, with the former more likely to be bespoke and built in-house, highly customised to a business model. The latter is more likely to lend itself to the creation of generic commercial solutions that can be sold off the shelf and applied broadly to a wide range of scenarios. Selection of these is determined by a firm’s appetite to spend, appetite for technology and innovation, and appetite to invest without immediate ROI.
Over the past few years, we’ve also seen that the big cloud providers are all starting to offer their B2B clients off-the-shelf AI capabilities (such as chatbots, NLP functionality, image processing software etc). These provide a relatively low cost route to adoption in the form of proof of concepts and pilots, with firms then moving on to more complex and expensive projects once the business benefits have been proven.
We also covered the five key pillars that underpin a firm’s ability to set themselves up for success with AI. The first is data – it’s the lifeblood of AI / ML, and must be complete, accurate and available in order to be successfully consumed and utilised by AI solutions. The second is technology capability – firms must have the necessary computing power, whether cloud-based on on-premises, accessible across their organisation so that engineers can quickly spin up sandbox environments for testing. Thirdly, firms need capable individuals such as data scientists, machine learning engineers, developers and production support, as well as the culture and mindset that’s needed in order to successfully develop and launch innovative AI solutions. Fourthly, firms also need strong supporting processes: procurement functions must enable tech teams to access and work with a range of smaller organisations, data governance must allow for data to be moved where it needs to be used. There must be processes for executing AI projects consistently, that allow them to be driven forwards and productionised successfully.
It is abundantly clear that AI solutions are heading towards a critical level of adoption, even where they are developed internally for a specific organisation and purpose. The key is developing a clear story and narrative about the longer term, more impactful return that can be secured through leveraging AI in the right way. Using readily available, low cost tools to build and run quick proof of concepts is a great, accessible, low-commitment start to that story for any organisation.