A great deal of time and energy is being put into figuring out the most effective way to leverage generative large language models in support of advisors and their firms.
And for good reason. Industry analysis after industry analysis has tools like ChatGPT positioned to reshape the way we work and redefine the term "modern advisor."
But is there any value in handing the reins of the robots over to the end client instead of limiting these tools to just being another part of an advisor's toolkit?
Researchers at the University of Chicago Booth School of Business, including doctoral student Alex G. Kim, and Chicago Booth professors Maximilian Muhn and Valeri Nikolaev, are looking at the feasibility of handing financial services customers.
In a working paper published earlier this year called "
Specifically, the team used the stock market as their laboratory and turned to GPT language models to summarize information disclosed by companies in annual reports and during conference calls.
In an interview with Financial Planning, Kim said the goal was to see if AI could trim the fat from disclosures and create focused summaries for investors interested in a particular topic, such as ESG activities or a firm's risk exposure.
Starting with a randomly selected sample of roughly 20% of all MD&As and conference call transcripts from 2009 to 2020, Kim and the others directed GPT-3.5 Turbo to generate a comprehensive summary of each document without referencing other documents or external sources.
There were questions about whether the shortened summaries could still deliver complete information. But Kim said the chatbot showed a great ability to preserve — and in some cases enhance — the value of the provided information, suggesting the presence of bloat in corporate disclosures.
Kim said the tools can also provide support when issues like jargon, lack of experience or complexity create hurdles that investors feel like they can't clear.
It knocks down a wall of intimidation between clients and the world of financial services. One that once removed may open a path to even more people working with advisors.
"There has been a lot of research and caution going on about readability or how informative these disclosures are, especially for retail investors. When investors are restrained by their cognitive constraints, they cannot fully process everything, and they tend to abandon some parts of the disclosures depending on their circumstances," Kim said. "If we can somehow distill this information and make it more relevant, easier to read and more accessible to retail investors, that would be very cool. Before, we didn't have the technology to do that. Theoretically, it was possible. But now, practically, we can do that."
Created by AI and research company OpenAI and released to the public in November 2022, ChatGPT is an AI-powered chatbot based on the GPT (Generative Pretrained Transformer) language model. With the help of deep learning techniques, the tool has the ability to generate conservational, human-like responses to text inputs.
The Chicago Booth study saw the tech produce MD&As and conference call transcripts typically less than a fifth of their original length. Sentiment was also amplified as documents were summarized, with positive documents yielding more positive summaries and vice versa.
Researchers say this may result from companies "hedging" their views with precautionary statements that don't contain much real information or are largely boilerplate. They also found that bloat tends to be higher when a firm reports losses, has negative sentiment and experiences negative stock market reactions.
These patterns are consistent with firms including irrelevant language in their disclosures to obfuscate negative information, according to a research brief.
Kim said when making the case for this kind of client-facing AI, especially amid the current wave of AI enthusiasm, a certain level of caution must always come with every ChatGPT implementation.
"It is a machine. It is not magic. Generally, what the machine is doing is just completing the sentences in a most natural manner. It's no more than that. When you ask the machines some problems that involve decision making, sometimes there is hallucination. It just gives you false information," Kim said. "If you try to correct it, it's going to say. 'I'm sorry,' and then it's going to make up another piece of information.
"We are minimizing the probability that it is hallucinating or creating facts. And then the decision making is on the investor's side."
Ken Lotocki, chief product officer of
Established in 2018, the Winnipeg, Manitoba-based firm relies on its proprietary, AI-based strategic advice manager to simplify the advice process by removing the trial and error that comes with building financial plans.
Lotocki said with safeguards against providing misleading information in place, these tools can provide contextual advice specific to the financial situation and literacy level of each user.
"For firms and advisors, it's been documented that advisors unfortunately spend far too much time in various financial software and less time supporting their clients. As an example, we've seen studies that it can take upwards of 17 hours to build a financial plan," Lotocki said in an email to Financial Planning. "Sticking with that example of planning software, generative AI/ChatGPT-powered tools can be leveraged within existing applications to enhance the user experience, provide streamlined access to existing and new features to assist advisors with UI navigation, analysis of unique situations and content generation such as report summaries."
He adds that Conquest has an ethos to democratize financial advice, and they believe generative AI can help make that happen.
"And when considering the end client, it's well-documented that we have a financial literacy problem around the world. This is a reason some clients avoid seeking financial advice more often from a professional," Lotocki said. "Clients are intimidated by what they don't know and can be embarrassed by their situation. This can be mitigated using generative AI/ChatGPT-powered tools, where the client at home can ask questions about their situation, ask questions about products and concepts that they may not understand, which can better prepare a client for interactions with a professional human advisor and firms."
Brian McLaughlin,
"We believe in introducing AI for advisors to empower them to streamline their operations and communications so they can focus on the most important matter at hand ‚—their clients," McLaughlin said, noting that Orion's third-party integrations allow advisors to vet all generated content first.
"The reality is that AI can make mistakes," he said.
McLaughlin also speaks in support of AI to handle tasks like comparing and contrasting portfolios, refining marketing content, responding to requests for proposals and more.
"We encourage advisors to challenge their internal teams to think of ways that they can leverage AI," McLaughlin said.
For Eduardo Fontes,
Fontes, who holds a master's degree in data science from Massachusetts Institute of Technology, said the proper constraint not only keeps AI-powered interactions safe from a legal standpoint, but the hyperspecificity can allow deeper responses related to the topic of interest.
"Because when you go in today on these large language models and you ask a general question, you probably don't want that brain to be thinking so generically and so widely on very specific documents. We have the ability to control that. That's No. 1," he said. "In my mind, the organizations that are going to be really moving forward with this technology, which I believe should be explored, will know how to control the ins and the outs. Because the ins and the outs do matter. Especially on the output side."
The average American investor is already taking to the tech. According to a recent consumer survey conducted by Magnifi, TIFIN's AI-powered marketplace for investments, 66% of Americans believe they understand how artificial intelligence works as a result of ChatGPT's presence.
In general, the survey finds that 60% of Americans are optimistic about artificial intelligence. Of that total, 24% said they are excited for it to make their lives easier and better, and 36% said that it seems like it has the potential to improve their lives.
The survey also found that 41% of respondents would make an investment decision based on guidance from AI. Millennials (57%) and Generation Z (52%) are the most eager to use ChatGPT for investment advice.
But a third (34%) of respondents said they do not understand artificial intelligence and are either unsure about it (24%) or don't trust it (13%).
