AI and the Future of Business Transformation in Financial Services
As artificial intelligence is becoming core to traditional financial services organizations no surprise: banks, investment management or leading financial advisory enterprises are in the accelerated path of researching solutions, to transform their entire businesses and innovate fast enough to provide smart, user-centric digital services that allow them to retain customers, build intelligent digital engagement and provide new revenue streams.
In order to understand some of the most actual financial market challenges & critical undertakings from the inside and both from the business & emerging technology viewpoints I decided to take a closer look into the finance community and to participate in the following events recently:
‘Artificial intelligence and the future of Financial Services’ organized by The Economist and ‘AI and Data Science in Trading’ organized by AIDST Digital Week, to interact with executives, decision makers and thought leaders from across the finance industry. Some of my random thoughts based on the events output you will find outlined below.
The need for external and more timely data has never been greater in the investment workflow. Last week I had an opportunity to participate in an AI & Data Science in Trading virtual event that brought together experts in the use of AI and advanced data analytic techniques within asset management, primarily for finding alpha, managing risk and optimizing portfolios.
When it comes to utilizing AI to optimize investment workflow I found ‘Alternative Data’ digital debate as well as ‘The reality of implementing AI within financial markets’ very insightful.
According to stats, nearly 73% of the everyday trading is executed by machines. Leading financial enterprises are investing in algorithmic trading because the level and volume of trade carried out by these machines is out of human bounds to process and execute.
These machines are based on a very complex model which takes into account the past historical financial data available as well as other information available on the Internet such as news — called alternative data. These systems take real-time trade decisions which maximize their returns.
Alternative data sets are information about a particular company that is published by sources outside of the company, which can provide unique and timely insights into investment opportunities. Alternative data sets can be compiled from various sources such as financial transactions, sensors, mobile devices, satellites, public records, and the internet. News and data companies as Bloomberg and Thomson Reuters now include alternative data in their offers.
Hedge funds and large asset management firms are gorging on “alternative data” to gain an investing edge. In the investing process, investors cannot just focus on fundamentals data, they need to understand what’s happening with ETFs (Exchange Traded Funds), need to understand online activity, what are people in social media saying (social-driven news feeds). Big Data is really the fact that data has just gotten massive and unstructured nature.
Now, from the AI/ML perspective, active research is going on in the field of stock trading, portfolio optimization, etc. Researchers are constantly trying to learn more and more information from the large volume of data available. A bit older models used only the numerical data available, but today’s system takes into account the financial news before it even reaches the humans and infers outcomes based on the news. In the future, we can expect machines to have a great control over the financial markets.
Manuela Veloso, Head of AI research, J.P. Morgan couldn’t be more accurate here, when it comes to the importance of an augmented look at the big picture, when thinking about harnessing human experience and the power of artificial intelligence to enhance financial services business process. This is what needs to be happening even more in the financial markets. Whether buy side or sell side or independent, the capabilities that can be unleashed when you bring together the best of machines (fast processing) and humans (creativity and decision making), the better your processes will be.
Terry Hickey, Chief Analytics Officer at CIBC shared a very interesting insight, that it’s relatively very simple to build a model to execute specific function, but to integrate that model into the existing workflow becomes very expensive when you’re dealing with large computing systems (banks, insurance companies, asset management and etc.). He predicts that in the next 2–3y time market will witness more integrations into financial backend systems and this trend will keep growing. Hickey added, that financial services organizations need to look at the ROI across identified business use cases, and unless one can show clear and measurable benefits, the money to fund AI projects will dry up.
Dave Oliver, Head of Nerve Center at RBS, said that the critical aspect of implementing innovation is the ease-of-use of AI and democratize the access on the enterprise-wide manner and Sally Eaves cited research from Edelman recognizing that consumer trust in business is at a 17-year low. As AI becomes more and more prominent in financial services, there’s a real possibility this could get worse. When people don’t understand AI or the inputs that drive a model — uncertainty grows. To combat this, enterprises should take two steps: make AI more transparent and explainable. It’s also very important to focus on diversity. The more consumers and users can understand technology, the more they’ll trust it.
When it comes to the reality of implementing Artificial Intelligence within financial markets, John Ashley (General Manager, Financial Services and Technology at NVIDIA) shared an interesting examples, outlined below:
- Scotiabank Use Case (deeply learning derivatives) about applying deep learning to derivatives valuation (an accelerated AI power risk model emulating whatever models you have, but you are moving a bulk of that compute time out of the trade window). Deep neural networks can be used to provide highly accurate derivatives valuations and these model compute valuations are approximately 1 million times faster than traditional Monte Carlo models.
- JP Morgan Deep Hedging Use Case. With deep hedging, machines can learn from large amounts of historical data to make more precise hedging decisions. Another advantage is that the technique allows for more automation of hedging, as machines can run parallel to identify appropriate hedges — this can make the entire process faster (already applied to index options books and can be expanded to more liquid vanilla products). Obtaining the optimal hedging strategy is a difficult problem, but representing the strategy as a neural network makes it tractable thus we’re more in control of risk mitigation.
Some of the questions / opportunities appearing in the context of next generation AI systems in the financial services I found really interesting during both events and for further exploration:
- Digitization of historical fundamental research and internal communication data as a source for alpha generation
- Automation & decision intelligence in the area of datasets evaluation / data providers / buyers.
- Human vs. machine intelligence ratio in the coming years when it comes to decision process in the portfolio management area
Ps Leveraging latest advancements in AI and emerging digital technologies will indisputably reshape landscape of highly regulated industries. What are still the biggest barriers to overcome? What are the most prominent opportunities ahead? I encourage you to interact via web, comments, or by getting in touch directly.