AI’s Evolution in Financial Services and Its Impact on the Future

Generative Artificial Intelligence (Gen AI) is a disruptive technology, taking over industries worldwide. While AI has been around for some time now, Gen AI is a newer type of AI that uses deep learning and neural networks to create its own content. With its ability to innovate, automate, and personalize outputs, it has the potential to immensely benefit the financial services industry.

We spoke with Paul Fahey, Head of Client Consulting and Channel Partnerships at Northern Trust, and Sandeep Varma, Co-Founder and CEO at Equity Data Science (EDS), about the evolution of AI in financial services and how they see it impacting the industry in the future.

How has AI evolved in the financial services industry?

Paul Fahey: Artificial intelligence has been around for many years, and its evolution has been nothing short of transformative. Developed back in the 1950’s, it was a rudimentary technology compared to what it is today. It started to pick up traction in the financial services industry in the 1980’s and ‘90’s with some of its first use cases including reducing the risk of human mistakes, fraud detection, predictive analysis and chatbots. Now, AI has grown from the simpler forms of machine learning to Gen AI and large language models (LLMs), which have the capacity to learn patterns and structures to create new content. AI is revolutionizing how businesses operate and how data is analyzed. It has paved the way for increased accuracy, efficiency and accessibility, setting the stage for the vast potential that Gen AI has in financial services.

Sandeep Varma: AI had been slow to adopt, until recently, when ChatGPT took the world by storm and introduced Gen AI. This latest technology boosted its use in a more meaningful way across all industries, including financial services. We’re seeing more use cases that provide valuable functionality to institutional investors beyond efficiency. While the efficiency use case is fairly clear in that it is helping our clients add structure to unstructured data, Gen AI also has the potential to augment analyst and portfolio manager workflows. Notably, it has brought greater efficiency and transparency to investment decision-making for fundamental managers, saving time and effort in their daily tasks.  

What has been the evolution of AI at Northern Trust?

Fahey: Northern Trust has been developing and leveraging AI for more than 10 years with rapid acceleration in the last five years. Northern Trust uses a variety of machine learning and AI solutions to process transactional activity more efficiently and mitigate risk. These include capabilities to digitize documents, reconcile accounts, and detect anomalous, fraudulent transactions. Northern Trust also conducts wide-ranging research and development on new technology, including Gen AI. This research includes using Gen AI to further enhance the document digitization process, optimize client responses, and develop independent agents to scale high-quality operational processes. Of particular note for custody services, we are extending our AI document digitization capabilities across a broad area of operational processes, from private equity to alternatives and collateral management.

What inspired Equity Data Science (EDS) to incorporate generative AI into its business model?

Varma: We have always been experts in quantitative data analysis but in recent years, as we have incorporated clients’ qualitative data in the form of research notes, Gen AI presented an ideal opportunity to structure the qualitative data and combine it with quantitative data (which by definition is already structured) to provide more efficiency and insights in their decision-making processes.

How do you foresee generative AI impacting investment data analysis in the future, and what specific benefits or opportunities do you expect it to bring to the investment industry?

Fahey: Gen AI has the potential to revolutionize investment data analysis. It has the capability to create advanced predictive models to determine plausible market outcomes, assess risks and opportunities and provide valuable insights. As a result, subject matter experts’ (SME) capacity will likely not be limited by the number of hours in a day. Much like the printing press enabled mass production of books and increased accessibility, Gen AI has the potential to significantly increase the capacity of SMEs.

Varma: Gen AI presents many opportunities for investment data analysis, which will significantly improve current processes and reduce the time and effort it takes to complete tasks. Ultimately, we will continue to see a logical combination of AI and human intelligence (HI), as there are some things you just can’t automate. The natural evolution is to adopt this AI + HI formula where humans leverage AI to maximize outcomes. For example, Gen AI has the potential to help investment analysts focus more on value-add tasks by reducing the time it takes to research and run through vast amounts of quantitative and qualitative data.

As generative AI models become more sophisticated, what challenges do you anticipate? For example, is there a potential for bias in the investment decision-making processes?

Fahey: While the potential of Gen AI in investment data science is vast, it does not come without its challenges. As more companies adopt Gen AI, there will be increased regulations around the technology which may pose a challenge. External inputs and outputs will also need to be properly managed. External inputs risk hallucinations, which are incorrect or false outputs. External outputs also risk exposing internal data, making data privacy a challenge. Humans in the loop (HITL) will play a key role in mitigating the risks inherent in Gen AI. Two key roles in the successful deployment of LLMs; writing effective prompts and interpreting the outputs, remain the domain of the human in the loop.

Varma: Gen AI does come with challenges, such as managing issues like hallucinations. Generating high-quality, synthetic data that accurately reflects real-world scenarios can be difficult, as Gen AI may unintentionally introduce biases or inaccuracies. Additionally, ethical concerns arise when using Gen AI, such as ensuring the responsible handling of synthetic data, addressing privacy implications and guarding against potential misuse.

EDS’s approach is centered on providing full transparency in the insights generated by Gen AI within the platform. We achieve this by referencing the exact quantitative and qualitative datasets used to form each response. The integration of AI with human intelligence (AI + HI) is set to become the industry standard, which will help to reduce concerns related to trust, accuracy and responsible use.

How is EDS using generative AI today? What are your future plans?

Varma: Within EDS, we are using Gen AI to:

• Automate ticker tagging to more thoroughly organize data
• Auto-tag research notes to the appropriate securities and themes
• Auto-tag with smart logic and sentiment tagging – i.e., “I want you to tag each note with one of these five tags. You determine which one goes to which note.”
• Run analytics on research data by asking questions – i.e., “How many broker meetings have I had with broker XYZ year to date?”
• Fuzzy search - type in a query and it'll bring up notes that are relevant to that query even if the specific words aren't in there
• Analyze and summarize portfolio risk and performance attribution

Investment teams will have the ability to ask the question “find me ideas that are interesting right now”. The results will come back with real insights to consider based on your database and offer the sources and relevant data to back it up.

Another idea for the future is to leverage Gen AI for marketing materials, such as an automated newsletter that could be very fund specific. Gen AI allows you to do somewhat complex activities at scale (GPT-type LLMs).

As generative AI continues to evolve, how do you plan to integrate the latest advancements and ensure that your clients benefit from cutting-edge capabilities?

Varma: As we deepen our understanding of real-world AI applications, we've recognized the potential of building specialized agents tailored to precise tasks, such as portfolio risk analysis and performance attribution. To optimize effectiveness, these agents are trained with data sets specific to each client, enabling a highly customized approach. Additionally, we leverage various LLMs, selecting those best suited to each unique use case to ensure precision, relevance, and maximum impact.

Fahey: The financial services industry has seen a remarkable evolution in AI technologies. It has revolutionized how businesses operate to meet their clients’ needs, and Northern Trust continues to research and invest in this technology to make processes more efficient. We are exploring interactive reporting solutions that go beyond simple chatbots to full, interactive financial reporting, enabling a level of conversational customization far beyond traditional portals and report generators. We are also bringing together our investment expertise to develop data analytics and insights to help clients optimize their investment portfolios much more rapidly in a fast-changing world.

The desire to incorporate Gen AI into business models will continue to grow due to its capacity to manage data, streamline content creation, enhance risk assessment and foster innovation. Successful execution will be heavily dependent on the development of the human skillsets and the quality and quantity of the data underpinning the models.