A big bet Morgan Stanley Wealth Management placed last year on a nascent technology involving a chat bot is paying off as the advisory arm of the firm last week revealed a partnership with OpenAI and its internal rollout of a refined version of ChatGPT as a tool for its advisors.

The AI tool will tap into the company's vast storehouse of research and data and make the information available through a chatbot platform to Morgan Stanley's financial advisors, according to the company.

“AI will transform a lot of industries, and I think this should lead to increased efficiencies for Morgan Stanley’s financial advisors that will help them personalize their client service,” said Michael Wong, a Morningstar financial sector analyst. “An advisor can do only so much personalization with 100 clients. The more time the advisor has, the more clients they can service. AI can give them that time.”

While Morgan Stanley is the first wealth management firm to harness this technology, if it works well for them, the firm’s competitors will follow suit, Wong said, adding that there still might be some wrinkles that need to be ironed out.

“As with any new technology, there are hurdles, especially around compliance,” he said. “They can train their AI application on only their research, but it’s an open question on what answers will come out, especially if someone asks a question that Morgan Stanley researchers haven’t formed an opinion on.”

But for now, the initiative puts Morgan Stanley, which has leaned heavily into internal-use technology as a way to improve the client experience, in a class by itself for about a year, as it could take a competitor that long to catch up, said Jeff McMillan, head of Morgan Stanley Wealth Management’s analytics, data and innovation group.

An Unexpected Beginning

The relationship with OpenAI, the San Francisco-based artificial intelligence research company that in the November unveiled its groundbreaking ChatGPT interface, began over a year ago when the tech firm crossed paths with Morgan Stanley’s technology specialists.

“We’re out in Silicon Valley with our own innovation team, which is always looking for disruptive technology,” McMillan said. “We were introduced to an unknown company at the time, OpenAI.”

A meeting had been proposed, he said, and in preparation for it McMillan took some publicly available speeches made by Morgan Stanley’s top brass—James Gorman, chairman and CEO, and Andy Saperstein, managing director and head of the wealth management division—to test the technology’s use.

“We put the speeches into the system and started asking it questions,” McMillan said. “It not just answered our questions, but it then made some critiques of the speeches that just blew us away.”

McMillan said he took an email he had written and put it through the system for a critique, and it identified some areas where he could have improved the text.

“As with any technology we’re looking at, we have to know that it’s real,” he said. “So we started using our own research and technology and got to the place where we absolutely believed this technology was real.”

Morgan Stanley’s early investment in OpenAI has resulted in the asset manager being the only financial institution that’s working directly with OpenAI’s software engineers, and in fact has a dedicated team supporting the relationship.

Morgan Stanley is working with GPT-4, OpenAI’s most advanced private system, to comb internally-used Morgan Stanley documents for information, not ChatGPT, which scours the entire internet and can’t help but reflect what it finds there, including erroneous information.

“We have a unique arrangement with OpenAI, where essentially we’re co-developing with them,” McMillan said. “Another firm would have to work with a consultant and another team of developers. So not only do we get a year’s head start, but if we have a problem we just pick up the phone and we’re talking with the people who created the technology.”

A Revolution For Advisors

For Morgan Stanley Wealth Management’s advisors and all the staff throughout the asset management arm, GPT-4 will be transformative, he said, as it will radically change the way information and solutions are accessed and tailored toward client needs. The rollout may be complete by late spring or early summer, he said.

Morgan Stanley is inputting its entire internal library into the project, including hundreds of thousands of pages housed across multiple internal sites. It’s from this global company content that advisors and other staff will be able to query GPT-4.

“A lot of criticism of the technology early on has been with some of the output people got from it,” McMillan said. “When you’re querying the entire internet, it’s impossible to know what is right and what is correct. We’re limiting the universe to content our experts have already created and that has been approved. There’s just an awful lot of it.”

GPT-4 needs to be trained and fine tuned by advisors through a process of asking questions and ensuring the answers are correct. And the advisors and staff need to be trained in the correct way to ask questions.

To that end, 300 Morgan Stanley advisors and support staff—including a cross-section of tenured advisors with at least two years in the business, and assistants and office staff throughout the business—are currently working with the GPT-4 system as a test group.

“It’s an incredible time save. So much of what an advisor and a team do every day is reactive in nature. Yes, a portion is ‘I’m going to do a plan, organize a meeting or rebalance a portfolio.’ But a lot is answering client questions,” McMillan said. “Some of those questions are routine, like how to do a wire, but some are not. Some questions you might have had only once in three months, and it can take 45 minutes or an hour to track down the correct answer.”

And in doing so, the person tasked with answering that question might also have to engage other departments, all of which increases the time impact of a single client question, such as “What are the top three tech stocks Morgan Stanley likes, give me a comparison of those three stocks and give me their benefits and risks.”

“In the old world, that would be three separate reports, plus the comparison and analysis,” McMillan said. “In the new world, that’s one question and an immediate answer. To be able to get an answer instantaneously is transformative.”

A New Skill: Prompting

The key for Morgan Stanley advisors to get a good answer lies in how well they learn to prompt GPT-4. Learning to prompt and talk to the machine is a skill, McMillan said.

“For example, in the old world, you might type ‘IRA rollover’ into the search bar, and then sift through all the links in the return until you find what you’re looking for,” he said. “In the new world, you would say, ‘I’m trying to do an IRA rollover for a 70-year-old client who is no longer employed.’”

GPT-4’s response would include instructions completing the rollover, justifications for doing so, a counterpoint if appropriate and links to important source documents.

For the employees currently working with the system, Morgan Stanley provided one-on-one training to explain how the platform works and how to create a good prompt, and McMillan said they would continue to do that, as well as to train staff on providing usable feedback so they can continue to fine-tune the system.

Above all, use of GPT-4 will free up significant advisor time that they could spend with their clients.

“Will lower-end tasks get improved and get more efficient? Absolutely,” McMillan said. “But in 2023, people still want to talk to another human being. The human connection of this is very real.”

He did not want to speculate on how or when Morgan Stanley might develop a client-facing version of the technology.

“There are probably hundreds of ways to leverage this internally and externally,” he said. “Right now what we’re doing takes lots of focus.”

One external area that has already benefited from GPT-4, however, is recruitment, he said.

“I’ve already been on recruitment meetings, sharing what we’re doing and how we’re doing it. In my mind there are a lot of reasons why people might want to be here, not just this,” he said. “But this is a thing no other firm has, and it’s really delivering value to our clients and our franchise.”