A Chatbot Powered By OpenAI Is Being Tested By Morgan Stanley's 16,000 Financial Advisors

To aid Morgan Stanley's army of financial advisors, the bank is launching an advanced chatbot based on OpenAI's latest technology.

‍

Jeff McMillan, head of analytics and data at the firm's wealth management division, tells Bankrate that the bank has been testing the artificial intelligence tool with 300 advisors and plans to roll it out widely in the coming months.

‍

Morgan Stanley's decision is one of the first announcements by a major financial institution after the success of OpenAI's ChatGPT, which went viral late last year by generating human-sounding responses to questions when people asked it. The bank is a juggernaut in the world of wealth management, with over $4.2 trillion in assets under management for its clients. There have been a lot of articles written about the promise and potential dangers of artificial intelligence for years, but it seems that mainstream users didn't fully understand the ramifications of the technology until after ChatGPT was introduced.

‍

McMillan says the tool is meant to help the bank's 16,000 advisors tap into the bank's enormous repository of research and data.

‍

“There is a desire for people to be as knowledgeable as the smartest person in our firm, McMillan said. "You have the feeling that you have our chief strategy officer sitting next to you while you speak to a client as if he were sitting beside you."

‍

The development of generative AI has dazzled users and sparked a race among technology giants to develop new products; however, some users have reported that it has led them in strange directions. Earlier this month, Morgan Stanley analysts shared their opinion that ChatGPT occasionally exhibits “hallucinations” and can generate answers that seem convincing but are actually incorrect.

‍

User guardrails

The tool will answer advisors' questions instantly, similar to ChatGPT. Nevertheless, it is based on GPT 4, which is a more advanced version of ChatGPT.

‍

Instead of generating responses from the entire contents of the internet, Morgan Stanley's tool only generates responses from the 100,000 or so pieces of research that Morgan Stanley has vetted for this use, so this tool should have fewer errors. In order to reduce mishaps even further, the bank has human beings checking the accuracy of responses, he said.  

‍

"We are trying to actually break the platform by testing it with humans," he said. "We are confident in the new tool as a result of high-quality information, better models, and ongoing monitoring processes," he said, adding that the bank has high expectations for its new tool.

‍

McMillan's move builds on its earlier efforts, including the introduction of machine-learning algorithms in 2018 that encourage advisors to speak with clients. Knowledge workers fear that technology may one day be able to eliminate them entirely with each new development.

‍

“Every industry will be disrupted in some way for what I will call routine, basic tasks,” McMillan said.

‍

Nevertheless, machines cannot replace people when it comes to catering to sophisticated clients, he said.

‍

“These things are not capable of empathetic understanding; they are simple mathematics that can regurgitate information, and they are able to do that very well,” he added.

‍