How Morgan Stanley is turning employees’ AI ideas into reality

If Jeff McMillan does his job right, it will look very different in three years.

“Think about it. We don’t have a head of PowerPoint at Morgan Stanley or Excel,” McMillan told B-17. “These are just enabling technologies,” he added.

He was named Morgan Stanley’s head of firmwide AI in March to help integrate the technology across the firm. While much of his job these days is focused on getting businesses up to speed with AI and implementing it efficiently across the bank, he said his ultimate goal is for the technology to be ingrained into workers’ everyday lives.

Since his promotion, McMillan has led the rollout of a few generative AI tools in the bank’s wealth-management division, and has more use cases in the pipeline, he said. The bank’s push into generative AI has been fueled by its early partnership with ChatGPT-maker, OpenAI, and coincides with Wall Street’s recent obsession with generative AI to boost productivity and reduce grunt work for workers.

McMillan encourages employees to pitch new AI solutions. His firmwide team acts as a filter and vets the ideas, which can come from practically anyone who’s done the required training at the bank. To avoid creating an unruly situation where thousands of technologists, analysts, and bankers are building their own AI tools, he’s devised a rigorous multi-step process that involves pitching solutions to some of the firm’s top execs and devising a business value proposition.

Jeff McMillan was the head of wealth-management tech until his promotion in March.

As part of his role, McMillan co-chairs an AI steering group formalized earlier this year, with Global Director of Research Katy Huberty. The steering committee, which has representatives from each department, vets all AI use cases pitched by employees.

The steering group is working through more than 30 use cases that are in various stages on the way to launch, McMillan said. AskResearch, an assistant that gives investment bankers, salespeople, and traders information found buried in tens of thousands of research reports, is the latest generative AI product to make it through the process since launching McMillan’s team.

Many of the pitches the steering committee sees fall into two buckets: use cases that are relevant to several groups, or that matter to a specific team or group of users. For the former, McMillan is able to coordinate teams across the firm to collaborate and build solutions together, with the aim of increasing reusability.

By structuring the AI approval process this way, McMillan hopes to enable the bank to innovate without sacrificing safety.

“While there might be creative tension between experimentation and process, I believe that a rigorous process will ultimately allow us to develop and deploy technology faster and more efficiently,” McMillan said.

Inside the 8-step process

Although pitching AI solutions is open to anyone at the firm, there is some leg work involved. Mainly, workers have to complete specific training on governance and AI principles and meet standards around information security.

The AI steering group meets every other week to listen to the pitches, usually going through five or six proposals. The steering group usually either approves or approves with conditions, like rethinking an aspect of the solution or coming together with other teams that pitched similar ideas. In some cases, pitches are rejected — something McMillan says he generally tries not to do.

“I don’t want to be in a position where I’m telling people no. I want to tell people yes, and this is the best way to get to it,” McMillan said.

For presentations that are approved, the next steps typically involve identifying the people who will execute and figuring out who from tech, legal, compliance, and risk needs to be involved. Workers going through this process also have to articulate deliverables and identify the risks, as well as having a plan for mitigating those risks. That might be a standard set of questions and answers used in testing or making certain teams aware of the potential risk.

They will also have to put together a business value proposition that outlines the benefits, which could be quantifiable, such as decreasing margin or operating costs, finding new revenue streams, or decreasing risk.

Every other week, the AI steering committee meets to review the status of these projects. At the end of the process, the group pitching presents a final time to the steering group for go-live approval to ensure all the conditions are met. Finally, use cases go into production.

“What we’re doing is we’re helping them prioritize. We’re grouping them, and then my team, we handhold you. We say, okay, what are you trying to do? We help you set up the environment. We make sure you’ve got the right level of APIs, we are by your side as you work with our legal, compliance, and risk process,” McMillan said of his business partners.

Similar Posts

Leave a Reply