Investment Banks and Machine Learning: Friend or Foe?

Investment Banks must start embracing machine learning to remain competitive. For instance, UBS is rethinking the bond issuance process through a Canadian tech startup – Overbond – which provides estimates for the timing and price of a potential issuance.

Investment banks advice companies in their assessment of capital needs. Serving as a bridge between the enterprise and investors, they recommend the best option to raise funds, being one of them the borrowing from the public through a bond issue. Among many hard decisions, they must help its clients decide on the volume and timing of the issuance, which includes the estimation of market demand. Although this task has remained unchanged for almost a century [1], UBS is now rethinking it through a Canadian tech startup – Overbond – which applies machine learning to connect all bond market participants (i.e. issuers, dealers, investors) in a digital platform, providing estimates for the timing and prices of a potential issuance [2]. UBS, as many other companies, is embracing technology with the aspiration of improving its process and enhance its value proposition. However, given that banks have positions themselves over time as sustainable and lucrative businesses, why it is important for them to even start thinking about technology? What approach should they take towards technology to stand out from the arising competition?

Arguably the most important factor for a client in the bond issuance process is the gauge of demand, as it defines the price and timing of the bond. However, investment banks execute this task by analyzing the liquidity of the bond market or simply having a chat with the asset manager 1. In other words, they are taking the most important decision based on the “experience” of the asset manager. Therefore, machine learning will drastically increase the customer experience, granting them the confidence of data-based decisions combined with deep market insights[3]. In this sense, artificial intelligence will enable the client to maximize the value of the transaction and make sure that he/she is not leaving money on the table. Therefore, it is crucial for UBS to start developing these capabilities, as machines and humans will be more powerful together than either is alone1.

Although it’s clear that UBS needs to be on top of cutting hedge technologies to remain competitive, there are several alternatives on how to approach this. The financial industry is facing more competition than ever, since numerous start-ups are arising as potential threats. Therefore, banks have basically two alternatives: treat them as enemies or allies. In this case, UBS chose the latter by establishing a win-win relationship with Overbond. UBS levered on its core capabilities and innovation-driven strategy to make this fintech start-up part of its business. From the start-up perspective, it is essential to have the support of a big bank to further develop the subjacent technology (i.e. machine learning), as it provides financial stability, strong networking opportunities and overall guidance.

Although applying machine learning in the bond marketplace is a key milestone for UBS, it is just a tiny step towards the infinite world of opportunities that this technology can provide. For example, they could roll out to other areas within the Investment Banking division. In 2016, J.P. Morgan introduces the Emerging Opportunities Engine, which helps identify client’s best positions for follow-on equity offerings through automated analysis of current financial positions, market conditions and historical data[4]. In addition, as described in Deloitte’s report, banks could leverage on machine learning insights throughout the whole M&A life-cycle process[5]. At an early stage, artificial intelligence enables a deep dive due diligence through a detailed analysis of the target company’s risks finding out the main issues to focus on. Secondly, analytics could make deal teams more effective throughout the negotiation process by providing insights on whether to pursue a specific target or move onto another one. Thirdly, machine learning could support the post-merger integration process by helping identify risks, synergies and other opportunities of value creation.

As financial institutions get involved at a fast pace into the technological world, there are several points to consider when evaluating these strategic decisions, from both the opportunities and limitations perspectives. First, since the value proposition of investment banks is relying more and more on data instead of expertise or regulatory knowledge, are we going to reach a point where these actors will become completely useless and companies will build these capabilities in-house? From the limitations perspective, James Manyika from the consulting firm McKinsey mentioned that one of the main practical issues of machine learning is data labeling and availability[6]. In this sense, is it foreseeable that technology companies and banks would agree to share their information to achieve a greater benefit through machine learning, especially in an era when information is money? Time will tell. (word count: 750)

Sources:

[1] “A Canadian startup applies machine-learning to corporate bond issuance”, The Economist, May 10th 2018, accessed November 2018, https://www.economist.com/finance-and-economics/2018/05/10/a-canadian-startup-applies-machine-learning-to-corporate-bond-issuance

[2] “Primary Bond Issuance: Digital Price Discovery”, UBS and Overbond, November 7th 2017, accessed November 2018, file:///C:/Users/socha/Downloads/overbond.pdf

[3] “Show me the case for AI in capital markets”, Accenture, November 2018, accessed November 2018, https://www.news.overbond.com/show-me-the-case-for-ai-in-capital-markets/

[4] “Machine learning is now used in Wall Street deal making, and bankers should probably be worried”, Business Insider, April 4th 2017, accessed November 2018, https://www.businessinsider.com/jpmorgan-using-machine-learning-in-investment-banking-2017-4

[5] “Not using analytics in M&A? You may be falling behind”, Deloitte, accessed November 2018, https://www2.deloitte.com/ca/en/pages/finance/articles/analytics-m-and-a-ia.html

[6] “The real-world potential and limitations of artificial intelligence”, Podcast McKinsey Quarterly, April 2018, accessed November 2018, https://www.mckinsey.com/featured-insights/artificial-intelligence/the-real-world-potential-and-limitations-of-artificial-intelligence

Previous:

Not just winging it: predicting airfare at KAYAK

Next:

NIVEA: Leveraging Open Innovation to Drive Product Development in a Race to be the World’s Leading Skincare Company

Student comments on Investment Banks and Machine Learning: Friend or Foe?

  1. This is a great article that clearly explains the numerous challenges that the investment banking world faces in terms of technological disruptions.

    Addressing your question, I think that investment banks will keep playing an important role in the issuance business. Banks will keep providing the means (could be either people or technological platforms) to issue financial products. However, I believe that a higher level of automatization will compress margins. Once machines incorporate the technical know-how necessary to connect all participants, companies will need fewer people, therefore reducing costs. According to Vikram Pandit, former Citibank chief, nearly 30% of banking positions will be replaced by technology [1]. In addition, the decrease in salaries expenses and the lower level of technical know-how necessary to access the issuance business will decrease barriers to entry. A higher level of competition and a reduction in costs will compress margins, forcing investment banks to think about developing new businesses to sustain their historical levels of profitability.

    To conclude, I agree with your argument that machine learning can either benefit or deteriorate the business model of investment banks. The ability of banks to adapt to the new circumstances will define their future.

    [1] “AI in banking: the reality behind the hype”, Financial Times, April 12th 2018, accessed November 2018, https://www.ft.com/content/b497a134-2d21-11e8-a34a-7e7563b0b0f4

  2. Super interesting read! While the application of this type of technology is still nascent, I think there is broad applicability to other aspects of the investment banking business, mostly as it relates to matching potential buyers and sellers of securities (i.e. bonds and equities).

    In response to your question about whether these actors will become completely useless, I don’t believe they will for a few reasons. First, the role of investment banks in the context of the use of this technology is as a broker-dealer. Regulators require companies to engage broker-dealers in order to sell securities to the public, and broker-dealers need licenses issued by the relevant legal bodies to perform this function. I think that overcoming this regulatory hurdle will be virtually impossible due to the implications around investor protections. In addition, the value of such a system seems skewed towards the issuer: predicting which investors will be interested in a particular bond issuance based on past purchasing behaviour can help issuers identify who to speak to regarding an upcoming issuance, but predicting when issuers are going to issue bonds is a far more complex task, given that it is at the discretion of managers and boards to decide how to fund their businesses and in most cases, this is fundamentally unpredictable. Investment banks provide investors with the value that this type of system lacks by bringing actual potential deals to them.

  3. Thank you – I thought the essay provided a good overview of potential areas of machine learning disruption within the context of investment banking. There are many sub-groups and divisions within an investment bank, each specializing in different activities and thus exposed to AI disruption at varying degree. While I agree that certain functions will inevitably be automated in the future (disproportionately back and middle-office functions), the role of an investment bank is unlikely to go away. Let’s take sell-side M&A investment banking for example. At the very core, the services that an investment bank provide in that case are i) executive relationships to top buyers in the relevant industries, and ii) sell-side advisory and negotiation that guarantees the seller the best deal. Both are very human-centered activities and cannot be easily replaced by machine algorithms.

  4. Could it be that many of the investment banking functions become automated by AI and the bank with the brightest AI becomes the winner who takes up the entire market? It may be that one of the big tech companies armed with a machine learning artificial intelligence decides to takeover the investment banking industry. Unlike any start ups, it has the resources to hire an entire team of investment bankers and replicate all the functions. No more 100 hour work week and financial modeling would be required. It may not replace all the bankers, but only a small fraction of the people currently employed in the industry would be needed to oversee the process. The investment banking as an industry will be fine, but the work (or job opportunities) might be affected drastically.

Leave a comment