Morgan Stanley wealth advisors are about to get an OpenAI-powered assistant to do their grunt work

While the company’s earlier efforts involved creating a ChatGPT-like service to help advisors navigate the firm’s reams of research, Debrief brings AI into direct contact with advisors’ most prized resource: their relationships with rich clients.

The program, built using OpenAI’s GPT-4, essentially sits in on client Zoom meetings, replacing the note-taking that advisors or junior employees have been doing by hand, according to Jeff McMillan, Morgan Stanley’s head of firmwide artificial intelligence.

“What we’re finding is that the quality and depth of the notes are just significantly better,” McMillan told CNBC. “The truth is, this does a better job of taking notes than the average human.”

Importantly, clients must consent to being recorded each time Debrief is used. Future versions will allow advisors to use the program on corporate devices during in-person meetings, said McMillan.

The rollout will serve as a real-world test for the vaunted productivity gains of generative AI, which took Wall Street by storm in recent months and has bolstered the value of chipmakers, tech giants and the broader U.S. stock market.

Morgan Stanley’s wealth management division hosts about 1 million Zoom calls a year, the bank told CNBC. While estimates vary, one Morgan Stanley advisor involved in the Debrief pilot said the program saves 30 minutes of work per meeting; advisors typically spend time after meetings creating notes and action plans to address client needs.

“As a financial adviser I’m doing four, five or six meetings a day,” said Don Whitehead, a Houston-based advisor who’s been testing the software. By “having the note-taking service built in through AI, you can really be invested in the meeting, you’re actually a lot more present.”

It remains to be seen what advisors will do with the hours reclaimed from essential grunt work. In a sense, Morgan Stanley’s projects in generative AI amount to a “grand experiment in productivity,” said McMillan.

If, as McMillan and others believe, advisors will spend more time serving clients and prospecting for new ones, the technology should boost Morgan Stanley’s growth in assets under management, as well as retention of clients and advisors.

Morgan Stanley’s wealth management division is one of the world’s largest with $5.5 trillion in client assets as of March; the firm wants to reach $10 trillion.

It will take at least a year to determine whether the technology is boosting advisor productivity, McMillan said.

“I’m the analytics guy, but the advisors will tell you that they’re at their best when they’re engaging” with clients, said McMillan. “None of them will tell you they love taking notes or looking at research reports, right? That’s not why they got into this business.”

Ultimately, Morgan Stanley’s vision for AI is creating a layer of technology that seamlessly helps advisors perform all of their tasks — sending proposals, balancing portfolios, creating reports — with simple prompts, Morgan Stanley wealth management head Jed Finn told investors in February.

Many of the core tasks set to be automated, such as parsing contracts and opening accounts, are universal throughout Morgan Stanley, including at trading and banking divisions, McMillan noted.

Finance jobs are among the most prone to displacement by AI, according to a recent Citigroup report. AI adoption could boost the industry’s profit by $170 billion by 2028, Citigroup said.

While the process is still in its infancy, McMillan acknowledged that business models will likely change in ways that are hard to predict.

“I think that there will be disruption in some areas,” he said. “We look back on all the things that we think we’re going to lose, but we don’t see what’s ahead.”

What’s ahead is the need for millions of prompt engineers to train AI to create the desired outcomes for companies, McMillan said; it took Morgan Stanley months to fine-tune prompts for Debrief, he noted.

McMillan said he even told his teenage children to consider careers as prompt engineers.

“They’re going to learn how to talk to machines, and tell those machines what to do, and engage with people and collaborate,” he said. “It’s a whole different game than how we’ve been doing work.”