How Generative AI is Transforming Chatbots in Financial Services

Enhancing client experience with AI assistants

AI and Machine Learning (ML) have made remarkable progress in various industries, and the financial services industry is no exception. With the generative AI boom brought on by Chat GPT, it’s possible that we are in the next wave of advances in the industry. One common implementation is of GenAI powered chatbots or AI assistants. Traditional chatbots are known to be clunky and hard to work with if you are looking for something deeper than an answer to a question like “help me get into my account.” Start asking it a more personalized question outside of the pre-determined script and you are likely to get frustrated with the answer.

Now, with large language models (LLMs), chatbots can be replaced with AI assistants giving users a more conversational experience and can handle questions outside of the range of typical questions.

Generative AI options for AI assistant customization

There are numerous generative AI model options available which means your firm won’t necessarily have to invest in developing its own individual model. To keep it in your own environment, you can use one of the various open-source models available like Falcon AI or Llama. Other models and open-source solutions can be found on Hugging Face as well. Additionally, your firm can choose to connect to a model through an API with your cloud service (for instance, Open AI has their GPT models available through Azure).

A model “out of the box” is typically not going to be accurate enough for the use case but there are options to tailor it to company specific data. One common method being used is called RAG (Retrieval-Augmented Generation). RAG is an AI framework that consists of extracting information, splitting it into smaller chunks, and storing the chunks in a library so that an AI Assistant can refer to the relevant information before giving the user an answer. With this technique, information can be added to the library as needed. The model can also be fine-tuned to the specific problem to help control the output if you want a certain voice, message, or writing style.

How can AI assistants help to educate your firm’s internal teams?

The Keiter Data Solutions team recently helped a client implement an internal facing AI Assistant solution using a RAG framework to help less experienced employees find answers related to products. Along with the answer, the source is also given so the employee can see where the information came from. Also, Morgan Stanley has been experimenting with GPT-3 and GPT-4 to create an internal facing AI assistant enabling personnel to quickly access relevant information spanning across thousands of documents.

In both internal scenarios, the AI assistant turns into more of a research partner. Jeff McMillan, Head of Analytics, Data & Innovation for Morgan Stanley has stated, “You essentially have the knowledge of the most knowledgeable person in Wealth Management—instantly. We believe that is a transformative capability for our company.” (1). The time efficiencies provided by an AI assistant tool could help your firm give better, more personalized service to your clients. It’s not seen as a tool to replace, but a tool to help enhance and improve service.

How can AI assistants help improve client experience?

For external facing applications, Bank of America has their AI assistant, Erica, which has had over 1.5 billion interactions since 2018 (2). In newer examples, there are AI assistants like DaveGPT for Aisera. Apparently, DaveGPT can auto-resolve up to 89% (3) of customer service requests. Also, Alpha from Publlic.com provides users with insights on historical and current market data. It’s not designed to give investment advice but aims to help consolidate information for the user.

In all scenarios, the tools haven’t removed the need for human interaction but created a more enhanced user experience (4). DaveGPT can’t handle all requests, but a customer is likely to get resolution before needing to wait for assistance. Also, Public.com stresses the importance of reviewing information and fact checking but it gives their users a place to start their research to draw their own conclusions.

AI risks – Look before your leap

Your financial services firm will need to carefully explore the data privacy and security risks in implementation of GenAI models. For instance, if your firm connects to a model through an API, what happens to the data, and will the data later be used to train the model? Remember: terms and conditions can change so constant awareness of sensitive data and how it’s being used is crucial.

Also, it is important to understand the current limitations of GenAI. As of today, humans still need to be involved in prompting questions and evaluating the output. Even if the model is robust and well trained with the appropriate guardrails, it’s still subject to bias and error. Your teams will need to think critically and evaluate both the prompt and the answer for accuracy.

Outlook

Generative AI promises exciting advancements across the financial services industry, particularly with AI assistants. The technology has great use cases and can be internal or external facing and it hasn’t removed the need for human interaction, just elevated the user experience. While the future holds great promise, it’s essential to approach this innovation thoughtfully, considering factors like data security, privacy, and ethical implications.

Interested in leveraging AI assistants for your financial services firm? Contact your Keiter Opportunity Advisor or the Keiter Data Solutions Team. Email | 804.747.000. We are ready to share our knowledge and ideas to help improve your internal team and client experience.

Sources

1. OpenAI - Morgan Stanley Customer Story
2. Bank of America Newsroom| Press Release July 13, 2023 | BofA’s Erica Surpasses 1.5 Billion Client Interactions, Totaling More Than 10 Million Hours of Conversations
3. Aisera | DaveGPT
4. Public | Meet Alpha - AI for Investors
5. Forbes | Will 2024 Be The Year That Generative AI Comes To Financial Services?