Machine Learning at the heart of JP Morgan’s growth strategy
Although markets have largely recovered from the 2008 financial crisis, significant increases in regulatory burden have affected the profitability of most, if not all, financial institutions. To cope with these rapidly changing regulations, the growing need for high operational efficiency, and the increasing sophistication of their clients, banks have been turning to new technologies like machine learning (“ML”)1.
JP Morgan (“JPM”), the biggest bank in the United States and one of the largest employers in the American banking sector, emerged from the financial crisis as one of few big winners yet its dominance is at risk unless it aggressively pursues new technologies and establishes technological leadership across financial services. To accomplish that, the JPM’s management will need to take advantage of “big data” and of all the information that banks have access to in order to create efficiencies and serve more clients with greater effectiveness, depth, and sophistication.
To that end, JPM has made it a company mandate to lead with technology innovation, investing $9.5 billion in 20162and planning to further increase its technology budget by $1.4 billion in 20183. The effort will be focused on streamlining resource-consuming processes and onboarding the right partners in the short term, and building internal capabilities in the longer term.
In the short term, JPM is expected to continue focusing on streamlining cumbersome, mainly back-office, processes that rely on pushing information from one place to another and onboarding partners that are expected to substantially improve the business. For example, in order to automate the daily routine and cut down the time needed to analyze the business correspondence, the bank has developed a proprietary ML algorithm called Contract Intelligence or COIN. The tool is now used to analyze the documentation and extract the important information from it. Applying this tool enabled the bank to process 12,000 credit agreements in seconds, instead of 360,000 man-hours. COIN is the latest bot to be launched by JPM, which also uses the technology to parse emails for employees, grant access to software systems, and handle common IT requests4.
In an effort to shorten the process improvement and new product development cycle, the company keeps tabs on 2,000 technology ventures, using about 100 in pilot programs that will eventually join the firm’s growing ecosystem of partners4. For instance, the bank’s ML software was built with Cloudera Inc., a software firm that JPM first encountered in 20092,4, while the bank recently made a strategic investment in n Volley.com, a startup that uses artificial intelligence (“AI”) to help large enterprises automatically generate training content for employees5.
In the longer term, JPM’s management will continue to invest in building in-house capabilities that will allow the reduction of the cost and the increase of the speed by which new products are developed and legacy processes are improved. Specifically, the firm has been setting up technology hubs for teams specializing in big data and ML to find new sources of revenue, while reducing expenses and risk. To help spur this effort, JPM is in the process of bringing in qualified data scientists who also understand how markets work and top academic researchers to lead internal R&D efforts. An illustrative example is Manuela Velasco, former head of Carnegie Mellon University’s Machine Learning Department, that was recently recruited to lead bank’s ML research team6.
However, while JPM has done an exemplary work embedding such technologies across business units, it also needs to assume responsibility for addressing the ethical concerns pertaining to these technological advancements. These include issues of fairness, safety, transparency, and accountability. Without systems compatible with these principles, the worry is that ML, and AI at large, will be biased, unfair, or lack transparency or accountability.
To ensure that AI ethics are taken seriously, JPM’s management need to undertake a series of steps. Specifically, they need to hire ethicists who work with bank’s management and software developers and develop a code of ethics that lays out how various issues related to AI will be handled. Responsible for the oversight will be an AI review board that will regularly convene to address such corporate ethical questions, while AI audit trails will help track how various coding decisions have been made. Lastly, AI training programs need to be developed so that staff operationalizes ethical considerations in their daily work.
In closing, with the increasing reliance of financial institutions on AI technologies to generate profits and increase market share, how can we make sure algorithms are fair, especially when they are privately owned by corporations, and not accessible to public scrutiny? Further to that, how can we balance the need for more accurate algorithms with the need for transparency towards people who are being affected by these algorithms? If necessary, are we willing to sacrifice accuracy for transparency, as Europe’s new General Data Protection Regulation may do?
(798 words)
Citations
1. Broughton, K. (2018). Leveraging AI and Machine Learning to Enhance the Customer Experience. AmericanBanker.com. [online] Available at: https://www.americanbanker.com/news/will-jpmorgans-splashy-tech-investment-pay-off
2. JPMorgan Chase Annual Report 2016(2017). JPMorgan Chase Annual Report 2016. JPMorganChase.com. Available at: https://www.jpmorganchase.com/corporate/investor-relations/document/2016-annualreport.pdf
3. JPMorgan Chase Annual Report 2017(2018). JPMorgan Chase Annual Report 2017. JPMorganChase.com. Available at: https://www.jpmorganchase.com/corporate/investor-relations/document/annualreport-2017.pdf
4. Son, H. (2017). JPMorgan Software Does in Seconds What Took Lawyers 360,000 Hours. Bloomberg.com. [online] Available at: https://www.bloomberg.com/news/articles/2017-02-28/jpmorgan-marshals-an-army-of-developers-to-automate-high-finance
5. Irrera, A. (2018). JPMorgan Chase invests in artificial intelligence startup Volley. Reuters.com. [online] Available at: https://www.reuters.com/article/us-jpmorgan-volley/jpmorgan-chase-invests-in-artificial-ijntelligence-startup-volley-idUSKBN1K721D
6. DiCamillo, N. (2018). JPMorgan recruits top academic for AI efforts. AmericanBanker.com. [online] Available at: https://www.americanbanker.com/news/carnegie-mellon-professor-to-lead-jpmorgan-chase-ai-research
Thanks for highlighting JPM’s blended internal/external source approach to ML platforms. I’d be really interested to learn more about how that process is currently being coordinated – specifically how these ventures are identified, monitored, and then on-boarded. JPM released a research paper on the applications of big data and ML in finance that opined data scientists with no financial experience or insight developing finance ML applications could be a risky proposition – will be interesting to see how (or if) JPM tries to follow its own guidance on that matter.
Interesting read. One thing that specifically jumped out at me was the hiring of Manuel Velasco to lead JPM’s machine learning research team. This draws strong parallels to Bridgewater’s hiring of IBM’s Ferrucci to head up their machine learning initiatives. It seems to me that at least in machine learning’s current form, financial institutions represent a premier employment opportunity for individuals focused on machine learning.
This raises further questions, in addition to the ones you posed above. Are there enough machine learning data scientists to go around? Is machine learning talent the true scarce resource when it comes to using machine learning to improve the operational efficiencies of an organization? If so, how would JPM change its HR initiatives in the coming years to make sure that it’s growth strategy doesn’t fall short due to shortages in employee supply?
Your point regarding the safety and fairness of the technology is well taken. One issue that the bank will have to grapple with is using machine learning to approve loans for customers. Safety, because JPMorgan will have to ensure that the information customers provide is safeguarded and not used in fraudulent transactions. Fairness, because a process that once included human judgement, the loan application process, would be fully taken over by machines. Balancing safety, fairness, and “keeping up with the times” by leveraging machine learning will be one interesting task for JPMorgan especially as the organization tries to retain its competitive advantage amongst global financial institutions.