Industry
Capital Markets
Unlocking Exponential Future Growth in Capital Market Firms
HIGHLIGHTS
- Wealth managers and financial controllers need data on a host of things—from the flow of funds and assets under management (AUM) to income, expenses, client profitability, and product pricing.
- However, the inability to gain seamless access to data and generate reports for in-depth analysis is a big challenge for many.
- The capability to handle queries in natural language will enable wealth managers to access information as required and interact with databases without the help of IT teams.
On this page
On this pageinpage
The business imperative
Quick decision-making and a broader view of data are critical to keep pace with the fast-changing wealth management industry.
Access to actionable, data-driven business and operational intelligence in real time is critical for astute business decisions. Wealth managers and financial controllers need data on a host of things—the inflow and outflow of funds in specific periods, assets under management (AUM), income, expenses, client profitability, product pricing, relationship manager attrition, customer churn, campaign management, and regulatory capital requirements. Similarly, financial controllers, top management, and wealth managers require a transparent view of the performance of key products and services in order to identify the most promising offerings as well as those whose performance is subpar. Such comprehensive data will help wealth firms define measures to boost their growth.
Wealth firms often struggle to make the right business decisions due to the lack of comprehensive and timely business intelligence, resulting in poor outcomes and dissatisfied customers. Gaining seamless access to data and then slicing and dicing it to generate reports for in-depth analysis require deep technical expertise. Business teams often lack the requisite technical skills to extract data spread across disparate databases, analyze it, and draw conclusions. While there are multiple self-service tools that can facilitate data extraction and analysis, to use them effectively, firms must equip teams with tool-specific technical skills or hire tech-savvy domain experts.
Consequently, business leaders are often forced to reach out to IT teams to access relevant data. These teams assess the request, identify relevant datasets, create data extraction programs, and deliver the information through a report. However, the turnaround on such requests varies from a few days to several weeks, delaying strategic business decisions such as how to deal with under-performing offerings—whether to withdraw them from the market or tweak them to better meet customer requirements—among others. This, in turn, results in unfavorable business outcomes, including loss of revenue and customer churn.
Enter GenAI
The solution lies in building the capability to handle queries in natural language.
This will enable wealth management business teams to quickly access the required information and interact with databases in natural language, thereby eliminating the dependence on IT teams. Compared with the existing cumbersome process, this approach will significantly enhance user experience and data accessibility, making data exploration and analysis more user-friendly, even for non-technical users.
Conversational SQL tools with the capability to handle natural language queries can be built by leveraging generative artificial intelligence (GenAI) and large language models (LLMs). For example, with a GenAI tool, wealth firms can post a query in natural language to instantly ascertain which relationship managers have an AUM below a certain threshold in a specific period—this intelligence can form the basis of new strategies to grow AUM.
Similarly, the tool can be used to gain insights on how products are performing in different markets. With such insights, firms can tweak existing offerings and design new offerings that better meet customer needs, resulting in new revenue streams. GenAI tools democratize access to knowledge and information, eliminating key person risk, enhancing information sharing across asset classes, facilitating easier access to documents and policies for support functions, and reducing time and cost spends on recreating old artefacts.
A vast playing field
GenAI promises a sea of potential, and more.
Several functions such as investment research, portfolio management, hyper-personalized strategies, client reporting, regulatory compliance, financial advisory, cybersecurity, and risk management can benefit significantly. GenAI-enabled research assistants can be deployed, forming a low-cost, ‘artificial’ army of research analysts and investment consultants in the financial advisory space. This will allow human analysts to focus on niche sets of data and expedite the overall research operation. Similarly, customer due diligence systems can be improved by integrating a GenAI-based chatbot into know your customer (KYC) processing for initial screening and document verification. However, business leaders in wealth firms are treading cautiously: they are adopting GenAI for internal purposes before using it in customer-facing functions.
GenAI can also help firms nimbly adapt to regulatory change. When the US announced stringent sanctions against Russia in response to the Ukraine conflict, restrictions were imposed on the international assets and remittances of Russian businessmen. Wealth firms had to comply with the sanctions while ensuring minimal negative impact on their business across the globe. This situation required well-coordinated, prompt action across various departments within firms. Large financial institutions and wealth firms scrambled to comply, setting up task forces to collect data, assess risk, and identify impacted client relationships. GenAI-driven self-analytical capabilities could have facilitated quicker turnaround in data analysis for such clients and helped relationship managers to proactively safeguard their business.
UK regulators have mandated climate risk disclosures in line with the Taskforce on Climate-Related Financial Disclosures (TCFD) framework. Compliance demands clear, comprehensive, high-quality information on the impacts of climate change, including risks and opportunities presented by rising temperatures, climate-related policy, and emerging technologies. Wealth firms had to review their engagements with clients and classify them into various climate change-related categories to comply with TCFD; for example, a customer in the mining sector is rated lower than one in the green energy area. Timely compliance became a challenge given the dependence on technical experts to extract relevant data for analysis and on functional experts to validate the extracted information. With the ability to learn from existing queries and generate data and reports in different formats, facilitate mining of information and in-depth analytics, GenAI can help address these challenges swiftly.
Ensuring compliance with the Sarbanes-Oxley Act (SOX) and the Markets in Financial Instruments Directive (MiFID) becomes easier with GenAI tools. The tool can seamlessly integrate with reporting and internal audit systems and pull reports needed to verify that the compliance software is working as intended, thereby avoiding unforeseen issues.
An NLP-based GenAI tool can help
Generally, wealth managers query KPIs for business decisions and reporting.
We propose a GenAI tool that leverages natural language processing (NLP) techniques to understand natural language queries on KPIs spanning total revenues, profit margin, gross sales for custody clients, and so on, and translate them into structured SQL commands that databases can process (see Figure 1). It can also perform tasks such as parsing, entity recognition, and sentiment analysis to accurately grasp user intent. Security concerns are addressed by applying role-based access controls using existing identity and access management (IAM) frameworks.
Figure 1: Proposed architecture of a GenAI tool for wealth managers
The tool’s ability to learn and respond to the specific business context of the application is an important aspect. Financial institutions can consider two approaches to enable this contextual learning:
In-context learning (ICL): ICL is a relatively easy and inexpensive approach to convert queries in natural language to SQL. The major benefit of ICL is that it avoids retraining of the LLM, which is a cost- and resource-intensive process, requiring huge sets of training data. The best part of this approach is that the model learns to formulate prompts from a very small set of contextual examples. Additionally, it is flexible, and can be done on the fly, based on the context. However, one challenge with this approach is that it is heavily dependent on the context which means that the prompt has to be extremely accurate to ensure high quality output.
Fine-tuning LLMs: This involves retraining a pre-trained LLM for a specific domain or application, which requires a large, labelled dataset, avoiding the need to train a model from scratch. An existing LLM is trained by leveraging NLP to convert it into SQL pairs using the firm’s structured data. The success of the model is heavily dependent on the quality and coverage of the training data. However, producing comprehensive and error-free training data is difficult. In addition, retraining the model is costly and time-consuming, requiring additional computational resources beyond those needed to run the tool.
Given the limitations of fine-tuning, we believe ICL offers an easier approach. By enabling the addition of contextual information extracted from a large number of historical queries received from business teams, financial controllers, and wealth managers, accurate prompts can be crafted. Chain-of-thought prompting can be used to train the model and guide its responses, resulting in precise and accurate data for business analytics.
And we rest our case: Learning from the best
Credit Suisse, a UBS company, is running a proof of concept (PoC) to determine the feasibility of implementing a GenAI-powered conversational SQL tool.
The tool is equipped with natural language capabilities to ease and speed up the process of accessing data for business decision-making. Given its favorable economics and superior results, we decided to adopt the ICL approach. The tool leveraged contextual business data such as KPIs, AUM, revenue, income, data dictionaries, schema mapping, among others, along with frequently used SQL queries to formulate the prompt. With this tool, Credit Suisse has already seen some early wins:
- Faster and better insights on KPIs
- Better self-service capabilities for business users
- Operational cost reduction of approximately $100,000 per annum
- Improved time-to-market for upgrades to existing products as well as new offerings
- Timely response to regulatory queries
- Rapid analysis of client profitability and performance of sales teams
In light of the prevailing geopolitical environment and the growing emphasis on sustainable return on investments, the demand for streamlined business decisions and prompt response is intense in the wealth management industry. GenAI can help accelerate decision-making and enhance productivity for wealth firms. Though ensuring accuracy, security, performance, and explainability will pose challenges, the rewards of business decisions backed by real-time intelligence are well worth the effort.
