Generative artificial intelligence (AI) could well be one of the most transformative technologies for the investment banking industry. Deloitte predicts that the top 14 global investment banks can boost their front-office productivity by as much as 27%–35% by using generative AI.1 This would result in additional revenue of US$3.5 million per front-office employee by 2026.
The allure of generative AI powered by transformer models2 has not escaped investment bankers’ attention. The potential of the technology to transform investment banking activities seems to be vast, and the applications are far-ranging.
AI and automation are not new to investment banking. In fact, machine learning/deep learning algorithms and natural language processing (NLP) techniques have been widely used for years to help automate trading, modernize risk management, and conduct investment research. However, despite the billions of dollars spent on automating the various functions across the transaction life cycle, there are still a fair number of tasks that are conducted using precious human capital.
But large language models (LLMs) could help automate many tasks, not only saving money but also improving worker productivity. It could also free up resources to spark innovation and enable front-office staff to focus more on productively interacting with clients.
Results of recent studies on generative AI’s impact on productivity look promising. One study by Stanford researchers found that generative AI boosted a call center’s productivity by 14%.3 Another study by Massachusetts Institute of Technology concluded that generative AI helped reduce time and improve the quality of work for marketers, consultants, and data analysts.4 One common finding is that the technology can level the playing field and can, in particular, assist lower-skilled employees improve their outputs and productivity. Nonetheless, initially, lower-skilled workers may need to exert greater validation efforts.
Given such promise, the industry is swarming with numerous proofs-of-concept (POCs) and experiments. JPMorgan Chase recently applied to trademark a product called “IndexGPT” that offers investment advice to customers.5 Wells Fargo is using LLMs to help determine what information clients must report to regulators and how they can improve their business processes.6
When Federal Reserve researchers evaluated GPT models’ ability to “decipher Fedspeak” (i.e., classify Federal Open Market Committee announcements as dovish or hawkish), they found that the algorithms not only were superior to other methods but also demonstrated reasoning abilities on par with humans.7 Several institutions are already using similar GPT models to analyze official statements and speeches produced by central banks.
Vendors to investment banks have also increased their investments in the new technology. Bloomberg recently launched “Bloomberg GPT,” a large language model built on 50 billion parameters and tailored for finance.8 Similarly, Pitchbook has a new tool called “VC Exit Predictor” that uses a machine learning algorithm to predict a startup’s potential growth prospects.9
Generative AI should be especially fruitful in areas where the output generation effort is high and validation is relatively easy.10 In the investment banking context, this capability can enable front-office employees to do their jobs better across a spectrum of activities, including marketing, sales, decision support, research, and trading, thereby boosting productivity. Professionals in these areas spend an enormous amount of time creating pitch books, industry reports, investment theses, performance summaries, due diligence reports, etc. Generative AI can help reduce the cost of content creation, enhance analytical capabilities, improve the electronification processes, and even reduce client call transfer rates.
Investment banks such as Goldman Sachs are also leveraging generative AI to help developers and coders create robust code more efficiently.11 Such competence is only expected to improve as these LLMs are trained on more parameters.
Our analysis suggests that the use of generative AI can boost productivity for front-office employees by as much as 27%–35% by 2026, after adjusting for inflation.12 This translates to an additional revenue of US$3 million to US$4 million per employee from an average of $11.3 million during 2020–22 (figure 1).
Productivity gains will likely vary by the inherent complexities of the underlying business. We estimate that gains will be the highest for the investment banking division (IBD), followed by equities, and then by FICC (fixed income, currencies, and commodities) trading.
The IBD, which includes equities and debt issuance, mergers and acquisition, and advisory, may benefit the most from generative AI, as it involves more repetitive tasks: We estimate that IBD productivity can be improved by an average of 34%. The technology can help generate initial deal structures and conduct due diligence, compliance, and valuation. In the areas of underwriting and issuance, generative AI can help with prospectus and term-sheet drafting and legal documentation.
Generative AI may also have a profound impact on trading. Automation and low-latency trading infrastructure have already morphed trading dramatically, possibly leading to greater market efficiencies, and reduced transaction costs.13 Traders leverage NLP and sentiment analysis to analyze markets, generate synthetic data for risk modeling, and optimize trading strategies. We estimate generative AI’s impact on such activities could significantly reduce time to understand market sentiment, catch anomalies, and place orders more easily and at greater scale.
In equities trading, generative AI can help traders quickly analyze, summarize company and industry fundamentals, run valuation models, conduct backtest trading strategies, and offer personalized trading recommendations to both institutional and retail clients.
FICC trading, on the other hand, often demands complex analysis and valuation, since it may also involve swaps/derivatives and a diverse array of trading strategies and risk parameters. Additionally, FICC markets tend to embody more systemic risk, so there is typically more regulatory scrutiny. While this offers space for generative AI to monitor bond yields, assess credit ratings, and provide real-time insights, the market-related uncertainty and volatility would require continuous validation from seasoned experts. These unique features may dampen productivity gains from generative AI, compared with equities trading.
The infusion of generative AI into the investment banking value chain will most likely come with potential legal, reputational, and other operational risks. It may also alter the dynamics with buy-side clients; as they also embrace this technology, the outputs they are able to generate with greater efficiencies may reduce dependency on the sell-side. Some clients may want to independently develop their own value streams and turn to banks only for the most high-value-adding services. Additionally, productivity gains could level the playing field by reducing barriers to entry, and further intensifying competition. But as investments needed to develop these LLMs become more substantial, this technology may also widen the gap among market participants and may put the smaller, boutique firms at a disadvantage.
Here are some key considerations investment banking leaders should explore when implementing generative AI into front-office functions:
Our forecast for productivity gains is based on the analysis of historical financial data from Tricumen. We then triangulated this data with the qualitative assessment of technology adoption trends and related efficiency gains. Using proprietary analysis, we projected the potential productivity gains across different investment banking tasks, also considering the relative complexity and varied impact of technology on fixed income currencies and commodities (FICC), equities, and IBD activities. The results were further validated using inputs from Deloitte subject matter experts to build a range of productivity gains.
View in ArticleFor more details on generative AI, please read “A new frontier in artificial intelligence” by the Deloitte AI Institute.
View in ArticleShana Lynch, “Will generative AI make you more productive at work? Yes, but only if you’re not already great at your job,” Stanford University, April 24, 2023.
View in ArticleShakked Noy and Whitney Zhang, “Experimental evidence on the productivity effects of generative artificial intelligence,” Massachusetts Institute of Technology, March 2, 2023.
View in ArticleHugh Son, “JPMorgan is developing a ChatGPT-like A.I. service that gives investment advice,” CNBC, May 25, 2023.
View in ArticleWilliam Shaw and Aisha S. Gani, “Wall Street banks are using AI to rewire the world of finance,” Bloomberg, June 1, 2023.
View in ArticleAnne Lundgaard Hansen and Sophia Kazinnik, “Can ChatGPT decipher Fedspeak?,” SSRN, June 7, 2023.
View in ArticleBloomberg, “Introducing BloombergGPT, Bloomberg’s 50-billion parameter large language model, purpose-built from scratch for finance,” press release, March 30, 2023.
View in ArticleLiveMint, “Can AI tools predict successful exits of startups? PitchBook has an answer,” March 31, 2023.
View in ArticleFor more details, please read: Deloitte AI Institute, Generative AI is all the rage, 2023.
View in ArticleRyan Browne, “Goldman Sachs is using ChatGPT-style A.I. in house to assist developers with writing code,” CNBC, March 23, 2023.
View in ArticleDeloitte Center for Financial Services’ analysis based on Tricumen data.
View in ArticleUS Securities and Exchange Commission, Staff report on algorithmic trading in U.S. capital markets, August 5, 2020.
View in ArticleFor more details, please read: Deloitte AI Institute, Proactive risk management in generative AI, 2023.
View in ArticleChristopher Savoie, “How quantum techniques could bring generative AI to the enterprise,” Forbes, April 10, 2023.
View in ArticleThe authors would like to thank the following Deloitte Center for Financial Services colleagues for their extensive contributions and support: Jim Eckenrode, Patricia Danielecki, Samia Hazuria, Karen Edelman, and Paul Kaiser.
Cover image by: Natalie Pfaff
