J.P. Morgan: Trades and payments with AI. What’s next?
J.P. Morgan is using AI to execute trades and assist corporate clients with payments. Other banks are also working on similar initiatives. Can AI help J.P. Morgan and other big banks retain customers and differentiate from competitors?
After experiencing the emergence of Fintech startups offering faster and customized solutions to clients, the Financial Services industry has started to see a trend of big companies strengthening their investments in technology and innovation. AI is one of the specific areas in which big banks have been increasingly announcing applications to improve client services and financial performance. After Retail, the Banking industry is the second biggest spender in AI, with an estimated amount of $3.3 billion to be invested worldwide in 2018 [1].
Front office activities are where big banks see the biggest potential for AI, but several of them are also starting to use it for data analytics, back office and middle office applications [2]. In a sector subject to regulation and with multiple tasks associated to required operating processes, AI also represents an opportunity to increase work capacity and allow employees to focus on higher value-added tasks. Automation and AI can free up to 10 to 25 percent of the work across functions, allowing people to spend more time on decision-making and other strategic and client-related activities [3].
AI and Machine Learning at J.P. Morgan
J.P. Morgan, one of the leading global financial services firm with assets of $2.6 trillion and worldwide operations [4], announced in 2017 the LOXM system, the bank’s new AI program that seeks to execute trades for clients with maximum speed and optimal prices, based on lessons learnt from billions of past trades [5]. LOXM is J.P. Morgan’s response to improved execution demands from its clients and has already shown benefits in performance in trials without raising risk management concerns, considering the system operates within the company’s risk framework. One of the possible evolutions of LOXM could be to train the machine to familiarize itself with the behaviors of individual clients in trading decisions [6].
Another concrete initiative that J.P. Morgan has announced is a virtual assistant for payments in its Treasury Services business, which comprises cash management, liquidity, trade and escrow solutions for clients. The assistant provides information such as balances and data from multiple accounts to corporate clients and the bank expects that it eventually learns their behaviors and can make recommendations [7]. J.P. Morgan already started a pilot program with clients in the technology, e-commerce and manufacturing spaces and the bank expects to increase the use of this application in a business segment that generated $7.6 billion in revenue in 2017 [8].
LOXM and the virtual assistant for payments are only two examples of J.P. Morgan’s efforts to increase service efficiency and effectiveness through technology, and through AI in particular. The bank is also starting to use AI, big data and machine learning to reduce risk, prevent fraud, enhance marketing and improve other services. New initiatives are expected to be announced considering that the firm has a budget of over $10.8 billion for technology in 2018 (with $5.0 billion dedicated to new investments) and that by now the company has assembled teams focused on innovations in AI, blockchain, big data, machine learning and bots [9].
What’s next?
Considering increased competition from smaller companies trying to provide financial services in more agile ways and with better quality, J.P. Morgan and other traditional banks need to continue investing in technology and developing partnerships to bring innovative solutions to their clients. AI and other initiatives are attracting clients’ attention and have the potential to contribute to retention and differentiation, two aspects that had been neglected by big banks in past years.
Additionally, increased efforts in processes streamlining through the use of automation and AI can help J.P. Morgan free up capacity that can be better used in client-related activities, which will have a concrete financial benefit for the bank.
With respect to the future of AI and its importance for J.P. Morgan several questions remain open. First, will AI and other initiatives be able to become key differentiators for the bank or considering the intense competition and investments from other banks will clients start demanding them as part of the basic expected services? Second, will clients trust technology enough to allow it to influence (or even make) relevant financial decisions? Third, how will existing and new startups react to the increased competition of traditional players such as J.P. Morgan, considering the significant difference in available resources that they have? Fourth, will processes streamlining reflect in job cuts?
(770 words)
[1] IDC. “Worldwide Spending on Cognitive and Artificial Intelligence Systems Will Grow to $19.1 Billion in 2018, According to New IDC Spending Guide”. March 2018. https://www.idc.com/getdoc.jsp?containerId=prUS43662418, accessed November 2018
[2] The Financial Times. “AI in banking: the reality behind the hype”. April 2018. https://www.ft.com/content/b497a134-2d21-11e8-a34a-7e7563b0b0f4, accessed November 2018
[3] McKinsey & Company. “The transformative power of automation in banking”. November 2017. https://www.mckinsey.com/industries/financial-services/our-insights/the-transformative-power-of-automation-in-banking, accessed November 2018
[4] JP Morgan Chase & Co. “Third-quarter 2018 results”. October 2018. https://www.jpmorganchase.com/corporate/investor-relations/document/e453a609-15bd-819d-fcfac560994d.pdf, accessed November 2018.
[5] The Financial Times. “JPMorgan develops robot to execute trades”. July 2017. https://www.ft.com/content/16b8ffb6-7161-11e7-aca6-c6bd07df1a3c, accessed November 2018.
[6] Business Insider. “JPMorgan takes AI use to the next level”. August 2017. https://www.businessinsider.com/jpmorgan-takes-ai-use-to-the-next-level-2017-8, accessed November 2018.
[7] CNBC. “JP Morgan is unleashing artificial intelligence on a business that moves $5 trillion for corporations every day”. June 2018. https://www.cnbc.com/2018/06/20/jp-morgan-is-unleashing-artificial-intelligence-on- -services.html, accessed November 2018.
[8] JP Morgan Chase & Co. “J.P. Morgan to pilot an A.I.-powered virtual assistant”. June 2018. https://www.jpmorgan.com/country/US/en/detail/1320568426081, accessed November 2018.
[9] JP Morgan Chase & Co. “JPMorgan Chase 2017 Annual Report”. April 2018. https://www.jpmorganchase.com/corporate/investor-relations/document/annualreport-2017.pdf?te=1&nl=dealbook&emc=edit_dk_20180406, accessed November 2018.
Thanks for highlighting this interesting topic. From my time working at a bank, machine learning was an important tool we utilized to stay competitive.
For example in FX trading, as an increasing percentage of the FX market moved to electronic platforms, it was crucial for Citi to price and execute electronic trades correctly, as narrow bid-offer spreads left little room for error. In this manner, we used machine learning to give clients the tightest bid-offer possible given hedging costs and market conditions. As the machine learned over time, with new data, our electronic pricing offering improved in tandem. Specifically, we utilized machine learning algorithms to analyze current market liquidity relative to averages and to determine the appropriate bid-offer spread adjustments for market moving events (such as new data). Additionally, algos allowed us to hedge more efficiently, skewing our pricing based on real time inventory needs.
Overall, for Citi, this allowed us to compete in the commoditized space of spot e-FX, while putting the customer first. However, we did not see this as innovation, rather as a necessity. Nevertheless, humans were still needed to make targeted business decisions, i.e do we need to take a loss on these trades to win bigger business?
From my experiences partnering with the buy side, clients trusted and found success utilizing algo trading strategies, capturing approximately 60% of major market trends. However, the issue was these funds became victims of their own success. After massive growth in 2016-2017, their AUM grew in proportion and their trades began to outstrip market liquidity. This was manageable in normal market conditions; however, in times of crisis, especially emerging market events, issues arose. What do you do when the model says to sell a large position and there’s no liquidity? For these decisions, humans are still needed.