Inside Morgan Stanley’s OpenAI Push
More financial advisers are using OpenAI-powered chatbot after Altman drama, but some say it's easier to pick up the phone and ask a human.
uring the most harrowing moments of the OpenAI leadership crisis, one of the startup’s biggest partnerships—a deal with Morgan Stanley that is supposed to demonstrate how AI can change the game at a giant, highly regulated Wall Street bank—was left twisting in the wind.
Executives at Morgan Stanley, one of OpenAI’s biggest customers, wondered what would happen to their chatbot for financial advisers if OpenAI fell apart. They discussed what they might do if Altman wasn’t reinstated as CEO and OpenAI employees left in droves, a person familiar with the matter said, though they didn’t lay out formal plans. There were no easy solutions. For instance, they discussed bringing the chatbot in-house, but that wound up being a nonstarter because of concerns about who owned the intellectual property and how it might run without OpenAI employees, the person said.
THE TAKEAWAY
• Morgan Stanley executives worried during Altman crisis about fate of adviser chatbot
• OpenAI adviser tool is used twice as much as Morgan Stanley’s prior, self-made chatbot; leadership crisis at OpenAI may have encouraged usage
• Some advisers at Morgan Stanley don’t find chatbot useful
But the chatbot never stopped working, and its usage among employees has actually shot up recently—at least partly because of OpenAI’s prevalence in the news, the person said. Detailed data on the chatbot’s usage couldn’t be obtained, but the number of searches employees are running is currently twice as high with the OpenAI tool as it was with Morgan Stanley’s previous virtual assistant, which the bank built in-house, the person said.
That increased usage will help improve the tool, which has so far gotten a mixed reception internally, according to several Morgan Stanley employees who spoke with The Information.
The chatbot, called AI @ Morgan Stanley Assistant, is designed to help financial advisers quickly pull up information for customers of Morgan Stanley’s wealth management business. After early trials with a small task force of wealth management employees, the bank rolled out the chatbot across the business in September. Executives have prodded workers to use it. Some employees have sung its praises, while others say it struggles to answer complex questions and they could more easily call someone internally who knows the answers.
A person familiar with the company’s AI operations said it has seemed like the tool isn’t widely used by financial advisers and is more often used by administrative staff.
Morgan Stanley declined to comment. OpenAI did not have a comment in time for publication. OpenAI executives including Altman have publicly thanked their partners for sticking with the company through the turbulence.
In early trials, advisers who spoke with The Information said they were unimpressed with the chatbot. The tool could easily pull up certain material—for example, the most recent Morgan Stanley research about Apple’s stock—but struggled to help with more sophisticated questions about, for instance, structuring an LLC under specific state guidelines. Advisers also had to learn to “talk” to the bot, rather than searching for keywords the way they would with a traditional search engine.
That’s a different scenario than Morgan Stanley and OpenAI painted in March, when an OpenAI webpage about their partnership quoted a Morgan Stanley executive saying: “Think of it as having our Chief Investment Strategist, Chief Global Economist, Global Equities Strategist, and every other analyst around the globe on call for every adviser, every day.”
The tool can only use the bank’s own internal data and research. That makes it a unique challenge for OpenAI, which typically can tap the internet for material.
The partnership between Morgan Stanley and OpenAI has wide-ranging implications. Altman himself brokered the deal, The Information has reported, and it is key to showing what ChatGPT can do within a heavily regulated industry with a huge employee base. The Information previously reported that even before Altman was fired, Morgan Stanley had planned to shift more of its AI software to rely on Microsoft’s Azure cloud, which hosts a version of OpenAI’s ChatGPT, and that Morgan Stanley had also been in contact with Anthropic before Altman’s ouster.
The OpenAI chatbot for wealth isn’t Morgan Stanley’s only artificial intelligence endeavor. The bank routinely meets with other AI providers, uses other companies for less ambitious AI initiatives and spent years working on the in-house chatbot prior to the partnership with OpenAI.
There is also an AI initiative called Project Genome that examines the financial DNA of customers of its wealth management business to predict what products and services they might need. If a customer has a young child, for instance, the analysis might suggest starting a college savings fund. If another customer has a lot of cash and isn’t keen on risky investments, it might suggest CD products with higher interest rates than basic savings products.
The bank also recently launched a product with TIFIN Group, a wealth management–focused AI company. That initiative, part of Morgan Stanley at Work, relates to donor-advised funds for charitable giving. Morgan Stanley at Work, which helps companies manage stock plans and other benefits for employees, is a key part of Morgan Stanley’s plan to grow wealth management revenue.
The company still plans to expand direct use of the OpenAI tool over the long term, from advisers to clients, a person familiar with the matter said. The bank needs to make sure it works out all the kinks before then. Over the much longer term, executives see extensive possibilities for using AI to improve productivity and performance across businesses beyond wealth management, including trading, investment banking, asset management and lending.
When Morgan Stanley banking analyst Betsy Graseck asked CEO James Gorman about the bank’s AI tools at an event in June, he said they were targeted more toward revenue generation than cutting headcount or saving costs.
“I would say this about the team there: They’ve got a much stronger technology bent than I had when I was running the business,” said the 65-year-old CEO, who is retiring next year. “I think it really has leapfrogged way past where my capabilities were.”