How can LLMs bring value to the financial industry?

Background

Artificial Intelligence (AI) and generative AI have been making headlines in the past few months. The interest was especially sparked by OpenAI’s ChatGPT, a conversational AI tool developed based on LLMs. LLMs have a wide range of applications and have shown promising results across a variety of use cases and industries. For instance, McKinsey reports that the productivity lift from generative AI can lead to an increase of 3-5% of annual revenue in the banking sector which is equivalent to $200 billion to $340 billion of additional annual revenue. In this article, we want to take a deep dive into these tools and see how they can be of value to the financial industry.

What are LLMs and where do they fall in the larger AI space?

A Large Language Model (LLM) is a type of AI model focusing on natural language understanding and generation. The fundamental principle behind them is the use of deep learning techniques, such as neural networks with millions or even billions of parameters, to analyse and learn from vast quantities of textual data. These models are trained on diverse datasets, allowing them to recognise patterns within text with much higher precision.

The term LLM is often heard with other phrases such as transformer models and neural networks. Understanding how LLMs fit in the larger AI space is helpful to get a better sense of their uses and applications.

In Figure 1 below, we show how commonly heard concepts such as AI, machine learning, LLMs, neural networks, and transformers fit in together.

Figure 1

What can LLMs do for the financial industry?

Given the large amount and variety of data that is available in the financial industry, LLMs can bring significant value-add to businesses in the sector. We explore some potential applications below:

1. Data-driven decision-making
   Given the potential to work with unstructured text data, LLMs are able to draw insights from data sources such as news reports, social media content, and publications. This allows companies in the financial industry to draw from novel and hitherto underutilised sources.
2. Optimising regulatory and compliance tasks
   LLM-based technologies can be used for tasks such as information retrieval and document analysis to assist with regulatory and compliance-related paperwork. LLMs are also able to automate monitoring and reporting tasks, allowing financial institutions to have pipelines that will function with minimal human intervention.
3. Customer interaction and support
   LLMs have boosted the capabilities and expectations we have around chatbots and virtual assistants. LLM-powered chatbots such as ChatGPT have shown an immense capacity for human-like communication experiences. Incorporating these chatbots into financial customer support services will improve the efficiency and the nature of customer interactions. For instance, a virtual personal adviser that can provide tailored insight into investments or personal financial management can be extremely well-received by customers.
4. Business innovation and efficiency
   We have recently seen a surge of LLM-based add-ons for existing tools and technologies. For instance, natural language-based instructions, programming assistants, and writing assistants are becoming extremely common. These LLM-based functionalities can bring about significant innovation and efficiency to the finance industry.

What does the future look like?

1. Increased efficiency: LLM-based technologies can automate and streamline a variety of tasks, especially activities such as content retrieval and generation, but also tasks such as programming. This will bring about significant and sector-wide efficiency increases.
2. Advanced decision-making: LLMs will enable companies to make better sense of data, particularly unstructured text data, thereby allowing for more informed decision-making.
3. Better customer experiences: With higher natural language processing capabilities led by LLMs, customer-oriented tools such as chatbots will be more capable of taking on a larger portion of customer support, as well as providing improved support services. This will improve the quality of customer experience while freeing up valuable human time and capacity to engage in more value-generating tasks.
4. Rapid changes led by innovation: The AI space is rapidly changing, with businesses constantly aiming to be at the cutting edge of innovation. Fueled by this, rapid changes can be expected in the financial sector.

How can businesses better prepare for LLM-led changes in the financial industry?

1. Embracing the changes and the challenges: AI is here for the long term, and will significantly change the way businesses function across the industry. Embracing the changes sooner than later will allow companies to make the most out of the emerging AI technologies. This is especially applicable in areas that contain tedious, process-oriented tasks. Using AI and LLMs to automate such tasks can create significant boosts in productivity.
2. Investing in research and development: The AI and LLM space is evolving rapidly, with new technologies and improvements introduced regularly. Companies that invest the time and effort to understand these changes and potential improvements are likely to gain a competitive advantage over time.
3. Upskilling the workforce: Enabling employees to go a few steps beyond their usual set of skills to building AI and LLM-based capacities will lead to increases in employee productivity. This will also future-proof the workforce, allowing businesses to confidently tackle the challenges that may emerge with rapidly changing AI.
4. Encouraging partnerships and collaborations: Partnering with research groups and other industry players in the space will help develop synergy to bring about the best possible outcomes. This will also enable firms to draw from the latest innovations in AI when improving their workflows.

How can evoML help?

evoML is a platform that brings the entire data science pipeline into a single place. With functionalities such as data pre-processing, feature engineering, model development and model evaluation all embedded into one tool, evoML enables data scientists to reach their machine learning tasks faster while allowing business leaders to derive better value from data available to them.

These are some key LLM-related functionalities offered by evoML that can level up a business in the finance industry:

1. Synthetic data generation from natural language instructions: The use of LLMs for synthetic data generation can mitigate numerous challenges in the finance sector, such as limited data availability, lack of representative data sample, and concerns emerging from data privacy. evoML offers some built-in natural language-based synthetic data generation features, which financial firms can use to supplement their existing data for modelling tasks.
2. Improving existing language models: evoML’s Model Hub feature enables users to upload their own language models to further fine-tune and manage using evoML functionalities. This will allow financial institutions to improve their custom models with relatively low effort, ultimately reducing the cost and time for deployment and maintenance of AI solutions.
3. High transparency and model code ownership for regulatory compliance: Financial firms, particularly banks, are required to maintain strict regulatory compliance. This complicates the process of implementing machine learning solutions, as these solutions may produce results that are not entirely explainable. evoML includes a range of metrics to evaluate model transparency and explainability, and gives the user the option to download and customise model code in order to ensure that models meet the necessary regulatory requirements.