Financial advisors say the technology will lead to big changes in investing, although customers still prefer a human touch for the time being, according to a survey of nearly 1,000 investors and financial advisors by Morgan Stanley’s wealth management arm that was published Wednesday.

A.I. will have big implications for finance, and 72% of investors call it a “game changer” in the making, according to the survey. “Most believe A.I. will be revolutionary for financial services,” the study said. 

“While A.I. is clearly groundbreaking, and we are just scratching the surface of its potential impact within financial services,” Jeff McMillan, head of analytics, data, and innovation at Morgan Stanley’s wealth management arm, said in a statement.

Even if A.I. will revolutionize the field, investors are unlikely to place all of their trust with the technology, and human advisors will continue playing a critical role, with more than 80% of survey respondents saying A.I. would never completely replace human guidance.

But investment firms that embrace A.I. will likely hold an edge over competitors soon enough, as 63% of investors said they would rather work with a company that leverages the technology versus a firm that doesn’t.

A.I. and investing

Morgan Stanley has tapped A.I. to enhance its wealth management arm for years. In 2017, the bank rolled out A.I.-powered software to assist its thousands of financial advisors in creating personalized plans. In 2019, it developed an A.I. model that could come up with original stock trading strategies based on patterns identified in market reports by Morgan Stanley’s own analysts. And in 2021, the bank announced a partnership with Microsoft to integrate the tech company’s cloud computing and A.I. tools with Morgan Stanley’s investment services.

McMillan framed Morgan Stanley’s big bets on A.I. as a “business requirement” while speaking at Fortune’s Brainstorm A.I. conference in 2021, saying that the bank’s use of A.I. to generate advice was “in the interest of the client.”

Morgan Stanley took another big step in March of this year, when it announced an investment and partnership with leading artificial intelligence startup OpenAI. Using the same large language model technology that powers ChatGPT, Morgan Stanley launched its own chatbot in a trial version that would help the bank’s 16,000 financial advisors navigate Morgan Stanley’s vast libraries of financial data and analysis to create personalized advice.

“People want to be as knowledgeable as the smartest person” at Morgan Stanley, McMillan said in an interview with CNBC at the time. “This is like having our chief strategy officer sitting next to you when you’re on the phone with a client.”

Morgan Stanley’s bet on A.I. assistants in its wealth management arm—which generated $6.6 billion in revenue during the first quarter of 2023—may be starting to pay off, according to its recent survey, which found that 74% of investors say financial advisors would work better if assisted by A.I.

But the survey’s findings also suggested that A.I. should only remain an assistant, rather than take on all of financial advisors’ duties, with 88% of respondents saying that face-to-face interactions with a human were “extremely important” for financial advisors.

While some A.I. models specifically designed to trade stock have shown early success, some experts have warned that investors should not blindly trust applications like ChatGPT with their money, as the technology is still prone to glaring errors. 

An April study from the University of Florida found that ChatGPT could outperform human analysis to predict stock movements based on financial news headlines, although in an interview with Fortune, the researchers cautioned that current models are still prone to making mistakes or misinterpreting information. 

They said that A.I. will likely become revolutionary for institutional investors like banks and large funds, but models like ChatGPT are only “a little better than tossing a coin” when it comes to helping individual investors make stock picks. While banks and large teams could run many iterations of different prompts and headlines to refine ChatGPT’s stock predictions and come up with an investment strategy, individual investors would likely have less success, as individual prompts with only a few headlines during the experiment only had a 51% accuracy rate.