While DTCC’s role as an industry-owned utility calls for risk mitigation and absolute protection of industry participants’ data, it also includes powering the financial services industry through technological innovations such as artificial intelligence (AI) and machine learning.
A future vision for the application of these technologies in the financial services industry was presented by industry thought leaders at the DTCC Fintech Symposium 2018 during two panel discussions.
Fuad Faridi, Associate Partner, McKinsey & Company, and Jeff Penney, Senior Advisor, McKinsey & Company, presented a future vision of capital markets using AI and machine learning. Ann Bergin, Managing Director and General Manager, DTCC Wealth Management Services (WMS), and David Kingland, CEO, Kingland, discussed how AI, machine learning and natural language processing can automate large datasets and identify patterns during the “Applying AI to Deliver Results” panel.
The Value of AI for Capital Markets
Discussing use cases being built today, Faridi and Penney talked about the value of AI, machine learning and natural language processing.
Taking a three- to five-year view, they presented a business case for a top 10 ranking investment bank, where leveraging AI and machine learning could potentially lead to a $700 million to $1 billion opportunity for revenue generation and cost reduction. Further, these technologies would provide improved regulatory compliance, early exception detection and increased speed and agility in serving clients.
Faridi emphasized the fact that AI and machine learning has come a long way since its overhyped days in the 1980s.
“The art of the possible has really shifted in this space, letting you do pretty amazing things,” Faridi said. “There is a real opportunity here for a step change in the capital market space which can shake up business models.”
The increases in computational power, the explosion of available data and the techniques employed to use this power and data were cited as the driving force behind these advancements.
Penney agreed, adding the acknowledgement that market making and trading activities are quite far ahead in the application of machine learning, natural language processing and AI. Part of these advancements has been due to necessity borne of the high volume of activity and data available.
“The complexity requires a machine solution,” Penney said, adding that automation shouldn’t be seen as a replacement of human judgment or experience. “We think about not replacing experienced people, but leveraging automation to create time for those people to focus on value-added activities and business decisions.”
Applying AI to Deliver Results
During his conversation with Bergin on developing and training AI engines for the financial services industry, one of the things Kingland emphasized was the importance of qualified personnel required to train AI engines.
“The idea is that cognitive systems are designed to extend what humans and machines can do,” Kingland said. “In order to develop these systems, you start with algorithms that are almost childlike in their capabilities. You have to teach them the nuances of the financial services industry. The technology becomes smarter as time goes on.”
Providing an example of AI application, Kingland and Bergin discussed DTCC’s Mutual Fund Services business, which has harnessed data mining technology and AI to enhance a key database that holds vast amounts of information on fund securities. This video provides a look into DTCC’s Mutual Fund Profile Security Database and how artificial intelligence continues to evolve.
The latest enhancement, a 2017 collaboration between DTCC and Kingland, uses AI to equip DTCC’s Mutual Fund Profile Service II databases to respond in real time as fund users input and update their data points. This prompts users of any exceptions that may occur, reduces the burden of data management and elevates the quality of data for clients.
The results of the enhancement, which began eight months ago, have been staggering. Bergin mentioned that there are now 27,000 securities in the database with five million data points and confidence rate of more than 99 percent.
“We are very enthusiastic about the future on this,” Bergin said. “We made the enhancement on an aggressive timeline and have seen great success with it. Today we have access to huge stores of complex data. Applying AI to those data sets provide insight and accurate analyses, allowing us to be more agile in our decision-making processes without the burden of data management.”
Looking Ahead with Deep Learning
Shedding some light on developments that are projected to take place over the next 12 to 24 months, Kingland said there are three types of humans that are engaged in training AI engines for this project: subject matter experts, engineers and analysts whose decision-making process is captured to train the AI engine.
Kingland also talked about AI engines that use a family of algorithms called Deep Learning, which consists of layers in a neural network. The output of one layer to input to the next, giving it the ability to learn without being explicitly programmed while adding higher accuracy.
“It is potentially very beneficial for users who want to extract and harvest regulatory grade data from any filing, anywhere in the world, in any language,” he said. “That’s very promising as we move ahead in the next couple of years.”