Read: The Need for Digital Transformation in Finance
women working at desk in office

Unlocking the Power of Generative AI in Financial Services Decision Making

Like most professionals, leaders in the financial services industry are exploring the business potential of generative AI (GenAI). Early signs point to an optimistic response—a recent Gartner survey found that 70% of financial service leaders believe GenAI will provide benefits within their organizations, rather than create risks.

Most FS leaders and their teams are in the pre-stages of GenAI adoption. “Only a small fraction of generative AI applications and investments have been disclosed today” by financial firms, Bain Capital reports. As we will find, only the largest, name-brand firms are taking formal steps to implement GenAI tools today.

Even so, off-record conversations suggest members of other firms are already using common GenAI tools—such as ChatGPT, Jasper AI, and others—as part of their work. But these publicly available GenAI tools for general queries and writing support only scratch the surface of what can be accomplished with GenAI in financial services.

In this article, we explore emerging capabilities, use cases, risks, and best practices for GenAI in financial services. With a focus on forecasting, predictive analysis, and decision support, we provide suggestions for how FS firms can approach GenAI safely and adopt new GenAI tools successfully in the future.

How GenAI Fits in Financial Services

Generative AI uses pre-trained machine learning models called Large Language Models (LLMs) to understand natural language and generate sophisticated replies. By letting a computer ‘learn’ from large amounts of data, In addition to responding to a query or prompt, GenAI can uncover hidden relationships between variables—such as risk factors or decision-making patterns—and make recommendations or even predictions about future outcomes.

GenAI functionality is accessible to virtually anyone by design. Historically, this differentiates it from most new categories of digital tools. (Consider the original learning curve for Microsoft Excel.) That’s because all generative AI tools are designed to respond to queries in a way that mimics human behavior: recognizing patterns, extrapolating from data, and making decisions.

There is clear potential for GenAI in financial services—where humans are responsible for making rapid decisions based on datasets and trends—with this in mind. As experts at Bain Capital describe, GenAI is a “sustaining technology”—rather than replace existing workflows, GenAI can integrate with existing tools, processes, and products in a way that is “generally accepted” by firms’ customers, not to mention their own staff. Also, “generative AI can fill the gaps within financial services organizations left unfulfilled by traditional AI” already in use today. 

Facing Compatibility Issues as the Market Grows predicts the financial services GenAI market will grow at a CAGR of 28.1% from 2022 to 2023, exceeding $9.4 billion in value during that time. But there are some compatibility issues FS firms and emerging GenAI solution providers must address before they formally adopt new, more sophisticated GenAI technologies.

Consider the regulatory environment of the FS industry. GenAI tools must be built in a way that complies with industry regulations—and this can be tricky to navigate. A GenAI tool may ‘think’ on its own, but firms are still responsible for any outcomes.

Fortunately, using GenAI in this context is comparable to existing investment and forecasting strategies that utilize qualitative elements such as news media and sentiment analysis. Indeed, GenAI can accelerate FS firms’ use of these resources, closing the gap between queries and insights that might otherwise require huge investigative efforts.

Traditional event-driven investment strategies and surveillance methodologies rely on mining for known behavior and patterns” in paradigms like event-driven news and social media sentiment analysis, AWS for Industries reports. While these “traditional event-driven investment strategies and surveillance methodologies rely on mining for known behavior and patterns… Generative AI has the potential to surface new themes and associated sentiment without direction.”

This strategic approach embodies the potentially “sustaining” aspect of GenAI, where its capabilities can enhance existing practices with greater speed and opportunities for deeper insight and exploration. It’s no wonder that the “forecasting & reporting: segment of GenAI growth in financial services represented the largest share of the application analysis market, according to

Emerging Use Cases for Dedicated GenAI Tools in Financial Services

Early use cases of formal GenAI tools at leading FS firms demonstrate their applications in existing workflows, sometimes supplementing or replacing existing AI functionalities. In early 2023, Goldman Sachs began “experimenting with generative AI use cases, like classification and categorization for millions of documents, including legal contracts,” AWS reports. “While traditional AI tools can help solve for these use cases, the organization sees an opportunity to use LLMs to take these processes to the next level.”

Meanwhile, JP Morgan adopted a chat tool for its teams to help them select the best investment plans for customers. Its IndexAI tool enhances financial advisors’ capabilities, helping them become more productive and provide more personalized services. In this way, GenAI chat is aiding employees as they carry out existing best practices.

Another approach involves “[training] one’s own domain-specific model from scratch,” as Harvard Business Review describes. Bloomberg recently launched BloombergGPT, which gives users access to Bloomberg’s substantial data resources via a natural-language interface comparable to ChatGPT. “Bloomberg’s data scientists employed 700 billion tokens, or about 350 billion words, 50 billion parameters, and 1.3 million hours of graphics processing unit time” as part of their efforts to improve knowledge management via GenAI.

Predictive Use Cases Are On the Horizon

Already, laypeople are using ChatGPT to pose questions about the stock market—from specific stocks and funds to more general concepts such as “What will be the best investment for 2021?” A predictive GenAI tool purpose-built for financial services can deliver a more sophisticated analysis and response to these types of natural-language queries.

For example, “LLMs can identify new trends in consumer behavior from social media content by clustering posts with similar meaning and assigning the clusters an aggregate measure of sentiment,” as AWS describes. GenAI can even analyze consumer sentiments associated with recent ad campaigns through publicly available data, helping firms predict stock performance and respond accordingly. The creation, acceptance, and formal adoption of these tools may be only one or two years away.

Roadblocks to the Use of GenAI in Financial Services

Even as the applications of GenAI in financial services become more sophisticated, firms must remain vigilant about how they use and deploy these tools. For example, it’s difficult to ensure GenAI models are in line with the latest regulatory changes and standards—especially in terms of how they use data.

Here is a closer look at potential roadblocks to GenAI financial services solutions development, acceptance, and adoption:

  • Inaccuracy: GenAI models may not always provide accurate output. FS firms and even GenAI application developers may wish to employ combinatory ML and AI techniques, such as ensemble models that combine multiple AI outputs to determine a reliable outcome. Human beings should fact-check any GenAI output as well.
  • Data Privacy and Security: Firms must be mindful of how their GenAI solutions use customer data and comply with applicable regulations. They should consider adopting secure ML models, such as federated learning, which ensures customer data remains protected.
  • Ethical Considerations: Popular GenAI tools have already demonstrated frightening instances of bias in their results—outputs that if taken at face value can lead to unethical decisions on the part of individuals and their firms. Strict governance policies can help FS firms ensure their final decisions are based on meaningful ethical standards and not blind trust in GenAI.
  • Misdirection: Depending on its design, a GenAI tool may provide confident-sounding recommendations without processing the most recent data before doing so. The human-like intonations of GenAI may obfuscate this fact, throwing firms off from accessing recent, unbiased resources and using their own judgment. GenAI can fall victim to a unique phenomenon called “hallucinations” as well, where bad results have already had legal ramifications for GenAI users.

These are just some of the challenges FS firms face as they work to adopt safe and successful GenAI tools in the future. With an understanding of both the potential benefits and risks of GenAI for financial services, firms can take steps to carefully evaluate new tools before investing.

5 Recommendations for Financial Firms

It’s likely that GenAI will soon play an important role in your own tech stack. Before you proceed, consider these recommendations for adopting GenAI solutions safely and effectively:

  1. Determine your most worthwhile use cases. Home in on the processes that have the most to gain, with the least possible risk. For example, consider instances where GenAI accelerates existing workflows without disrupting existing ethical standards and best practices.
  2. Investigate data privacy regulations associated with your efforts. Study the legal implications of deploying GenAI solutions in your specific region or field. Ensure any use of customer data complies with applicable laws. Review these factors every time you consider a new GenAI solution.
  3. Develop an ethical code of conduct. Establish a clear set of guidelines regarding how and when it is appropriate to use GenAI-generated outputs. Describe to employees the instances, workflows, and tools where these guidelines take effect; consider creating a disclosure for your customers as well.
  4. Test extensively before deployment. Ensure any new GenAI tool produces accurate output in a variety of scenarios and meets appropriate regulatory standards before rolling it out to your teams or customers.
  5. Monitor results regularly. Create appropriate standards and implement extensive monitoring shortly after each GenAI tool is implemented. Conduct random QA checks throughout the year to ensure your solutions are delivering the right results.

GenAI is a Building Block

Over time, GenAI will become as ubiquitous as Word processing. But like Word processing, it is only one part of a digital menagerie of financial services tools. We must remind ourselves that GenAI is not an intelligence of rationality and choice; it ‘predicts’ the right words in response to our queries, rather than ‘creates’ those words based on personal insights and experience. Financial services professionals who keep this in mind will be in a better position to incorporate GenAI solutions safely and effectively.