Use cases depend on the ROI from driving operational efficiency or generating revenue.
This article was first published as Trading Reimagined in Traders Magazine. Trading Reimagined is a biweekly content series that examines how the transformative power of technology is prompting a reimagining of the markets. Trading Reimagined is sponsored by Exegy.
There are many potential use cases where generative AI solutions could add significant value, according to consultancy Accenture.
Matt Long, senior managing director, global capital markets lead at Accenture, told Markets Media that clients are in serious discussions about using generative AI and the capital markets industry is moving quickly. Generative AI replicates human capabilities as it can read, comprehend, reason, make suggestions and create new content — such as computer code, text and images — based on a model that is trained on a huge dataset, while using feedback from human users to support continual learning and reinforcement.
Long said: “Use cases depend on the business case, or return on investment, that is going to drive operational efficiency or revenue generation.”
Capital markets firms either own or have access to a vast pool of data and Long said in a blog they can use generative AI to analyze this data and produce insights about their customers’ behavior which may provide opportunities to cross-sell or provide new products. Firms will be able to generate very tailored content that allows proactive outreach based factors such as sentiment analysis or credit market risks, and will be able to monitor changes that might affect an individual firm, credit score, or geographic activity.
Matthew Long, Accenture
“Clients absolutely say generative AI is a priority,” added Long. “Most people I talk to are very enthusiastic about the opportunity.”
Accenture said there are also opportunities for capital markets to improve customer focus for institutions by tailoring reporting portals and providing them with a unified view of their activities on a timely basis. There are also opportunities to simplify and improve the experience of client onboarding and Long said there is a lot of focus on know-your-customer and due diligence.
“In terms of enhancing customer focus, generative AI opens the way to provide every individual client with a customized and hyper-personalized experience, while empowering relationship managers and advisors to focus on value-adding activities and client needs,” added Long. “The technology can also help to improve and streamline customer interactions and communications through intuitive, natural-language chatbots that provide a fast, responsive and convenient service.”
Capital markets firms also typically have large technology teams to produce, maintain and deploy code and AI can help make these processes more efficient.
“There is now something called AI Ops to drive efficiency around technology production and speed of production,” Long said.
He expects slower adoption of generative AI in the middle and back office due to the need to change legacy technology, even though these areas have the most significant opportunities to improve operational efficiency. For example, in some cases up to half of operations activities happen outside core systems.
“The initial use cases in operations will be around work orchestration,” added Long. “It is a quick win if firms can understand what is going on in email traffic and then proactively orchestrate some of the allocation of work activities.”
There is also the potential to enhance operational productivity by using generative AI to automate and ‘bot-shore’ operational processes while giving humans smart tooling to help them do their jobs better and faster, and monitor the quality of results.
One of the main hurdles for firms in using generative AI around data quality and disparity.
Long asked: “How do you get consistent, complete data when you are dealing with an industry that typically has inconsistent data scattered across multiple systems and in very siloed business lines ?’
Another challenge for capital markets firms is having to operate within a highly regulated environment that is constantly changing. Long highlighted that AI can help rationalize and optimize controls, automatically track regulatory changes and respond by modifying existing controls or developing new ones. Use cases will also be subject to the framework developed by regulators.
“We see a very big focus across all industries, but certainly in financial services, around what we call Total Enterprise Reinvention,” said Long. “Every winner of tomorrow is going to require modernisation around a digital core and know their strategy around cloud, data and AI.”
Use case – Schroders
Charlotte Wood, head of innovation and fintech alliances at Schroders, said in a blog that the UK asset manager is very excited about generative AI giving people faster and easier access to information that already exists.
For example, investment teams have to consume a lot of data every day, and humans are limited in how much data they can usefully factor into investment decisions.
Charlotte Wood, Schroders
“AI presents an opportunity to vastly improve data consumption and application, thereby improving investment decisions and client outcomes,” she added.
The asset manager believes that real value lies in using generative AI on Schroders-specific data and the asset manager has built an internal version of ChatGPT with some extra functionality.
“Importantly though, the data from this technology is not being fed into the internet, so it’s not public information,” Wood added. “It’s being stored entirely by Schroders.”
Alex Tedder, head of global and thematic equities at Schroders said in the blog that financial markets are particularly excited about the application of generative AI to businesses and the productivity gains that can be realized. He estimated that the annual addressable market could be around $450bn, even on a conservative basis.
“These are big numbers, and that is without productivity gains,” added Tedder. “The potential for cost savings and productivity gains is significant, and that’s why financial market participants are very excited.”
Alex Tedder, Schroders
Tedder continued that financial markets have been very efficient in pricing the potential impact of AI, particularly with regards to revenue growth in the software and semiconductor sectors.
“What the market hasn’t really done yet is taken a step back and thought about what it will mean in terms of adding value in other sectors or industries,” he said. “And at the corporate level, this is where it gets very interesting.”
Use Case – London Stock Exchange Group
Prompt engineering using financial text data in GPT models can improve performance for sentiment and theme classification according to a report from LSEG Analytics, ‘Using GPT-4 with prompt engineering for financial industry tasks.’
The report in May 2023 said: “Using prompt engineering for sentiment classification, GPT-4’s performance was seen to further improve, indicating that prompt engineering is a valuable area for performance optimisation.”
LSEG said it uses large language model capabilities in a variety of products including summarisation, entity recognition, topic detection and sentiment analysis. For example, SentiMine provides sentiment analysis of earnings/ conference call transcripts for a range of financially significant themes to help analysts find the relevant updates needed to make a better-informed decision.
The study found that the latest GPT models demonstrate clear performance improvements over existing models for sentiment and theme classification, but LSEG believes there are clear opportunities for using GPT far beyond these tasks.
“The pace of updates to GPT and the performance improvements of GPT-4 are captivating for product implementation; we look forward to discovering new opportunities in this fast-growing field,” added the report.
However, Carl Carrie, head of analytics strategy at LSEG, warned in a report that markets manipulation via deepfakes is a particular risk in the financial services industry.
Source: LSEG
Carrie said in a study, ‘New AI Risks with DeepFace and DeepFakes’, that financial companies must be aware of these risks and take steps to mitigate them through a combination of regulatory oversight, ethical guidelines, and transparency in the development and deployment of these technologies.
“Mitigation will over time likely need to include investment in advanced detection and prevention technologies, education and training for employees, and the development of ethical guidelines and regulatory oversight to ensure responsible use of these technologies,” Carrie added.
