Morgan Stanley moves forward on homegrown AI

The new tool, AI @ Morgan Stanley Debrief, which summarizes video meetings and generates drafts of follow-up emails based on them, was Morgan Stanley’s second generative AI use case built in collaboration with OpenAI. It follows the September 2023 rollout of its AI knowledge assistant tool that helps financial advisers quickly track down information from Morgan Stanley research.

Morgan Stanley’s approach to working directly with model builders rather than just grabbing off-the-shelf technology is becoming more common across the financial services industry, said Alexandra Mousavizadeh, co-founder and chief executive of Evident, which tracks AI adoption across the sector.

With a build approach, “you get to design it, make it to fit into your workflows and just make it more seamless," Mousavizadeh said. Regarding tools like Morgan Stanley’s, she said, “The benefit of this app for a bank is that it’s able to work with the data inside the bank and fit seamlessly into the workflow of the financial analyst. No doubt that this type of app will be incredibly useful in other sectors too," citing insurance and retail as examples.

Other examples include BNP Paribas, which this month announced it was in a partnership with model builder Mistral AI. TD Bank also said this month it was working with Cohere. Mousavizadeh said that the majority of banks are working with several model-makers on the market, including Meta Platforms’ Llama, Anthropic’s Claude, Google’s Gemini and OpenAI’s GPT.

Wall Street Journal owner News Corp has a content-licensing partnership with OpenAI.

The companies are certainly experimenting with off-the-shelf AI tools as well, she added. Goldman Sachs chose to deploy Microsoft’s GitHub Copilot across the firm as its generative AI coding tool. It also created an AI platform that its developers are using to build custom applications on top of models such as Llama, GPT and Gemini.

Both of Morgan Stanley’s tools were built using OpenAI’s GPT model and in consultation with OpenAI, as part of the strategic partnership the two companies signed in early 2022. Morgan Stanley said that after assessing the product landscape for this use case, it opted for customization and seamless integration afforded by building the tools in house with OpenAI’s technology, rather than buying fully built solutions off the market.

OpenAI Chief Operating Officer Brad Lightcap said, “Working closely with Morgan Stanley has been a great opportunity to build and deploy AI that meets the needs of a highly regulated industry."

Technology providers increasingly offer kitted-out AI premium products, although they have yet to gain traction among many enterprise customers. Tools like Copilot for Microsoft 365 or Gemini for Google Workspace are turning out to require a lot of hand-holding to make them work in the enterprise, leading some CIOs to question the actual returns they offer against their high price tags.

At the same time, offerings from cloud software vendors such as Workday and Salesforce aren’t resonating in the market yet, leading to disappointing earnings, in part, analysts say, because enterprises are leaning toward building custom applications.

With Morgan Stanley’s meeting summarization tool, “some people would argue, well, you can use Zoom to summarize it. Or you could use Teams to summarize it. That’s true," said David Wu, Morgan Stanley’s firmwide AI head of product & architecture strategy, but he added, “Even if the product may exist, there’s advantages of building and customizing that solve our user needs better."

For example, Morgan Stanley financial advisers typically use Microsoft Outlook for email, Zoom for meetings and Salesforce to keep notes from the meetings. By customizing a homegrown tool, the company was able to integrate with all of those systems.

Wu said that as companies such as Microsoft and Salesforce continue to consider how their tools integrate with each other, Morgan Stanley is regularly assessing its approach. Wu noted that buying has typically been the company’s modus operandi in the past. But he said, based on the current level of maturity of out-of-the-box offerings, building makes more sense.

“We figured, let’s go in and insert the AI into the workflow where people already live, versus trying to buy a product and retrain everyone on a new tool that, by the way, may or may not connect to another tool that they’re using," Wu said.

Morgan Stanley says it is too early to share results of the meeting summarization tool. As for its knowledge assistant tool, it said 98% of teams are using it regularly but declined to share more detail on the financial returns of the tool.

Write to Isabelle Bousquette at isabelle.bousquette@wsj.com