Jeff McMillan is head of enterprise AI at Morgan Stanley. His team vets all new AI proposals, but business leaders ultimately decide what to use. He outlines the multiple steps from proposing an idea to deploying it in production.
If Geoff McMillan does his job right, things will be very different in three years.
“Think about it: Morgan Stanley doesn’t have a head of PowerPoint and they don’t have a head of Excel,” McMillan told Business Insider. “These are just enabling technologies,” he added.
He was named head of enterprise AI at Morgan Stanley in March, where he helped integrate the technology across the company. While much of his recent work has focused on bringing AI to the enterprise and deploying it efficiently across the bank, his ultimate goal is to enable technology to improve the daily lives of employees. He said that it is important to penetrate the society.
Since his promotion, McMillan has led the rollout of several generative AI tools in the bank’s asset management division, and more use cases are in the pipeline, he said. The bank’s generative AI efforts are fueled by an early partnership with OpenAI, the creator of ChatGPT, which Wall Street recently announced to improve productivity and reduce drudgery for employees. It aligns with my enthusiasm for generative AI.
McMillan encourages employees to pitch new AI solutions. His company-wide team acts as a filter, vetting ideas that come from just about anyone who has gone through the necessary training at the bank. To avoid creating an unwieldy situation with thousands of engineers, analysts, and bankers building their own AI tools, he pitched his solution to several of the company’s top executives. , devised a rigorous multi-step process that included devising a business value proposition.
Jeff McMillan was head of wealth management technology until his promotion in March. Provided by Morgan Stanley
As part of his role, McMillan co-chairs the AI Steering Group, which was formally established earlier this year, with global director of research Katie Huberty. A steering committee made up of representatives from each department reviews all AI use cases proposed by employees.
Related articles
The steering group is working on more than 30 use cases in various stages of launch, McMillan said. AskResearch, an assistant that provides investment bankers, salespeople, and traders with information buried in tens of thousands of research reports, is the latest generative AI product to go through the process since McMillan’s team launched it. .
Many of the pitches the steering committee considers fall into two categories: use cases that are relevant to multiple groups, or use cases that are important to a specific team or group of users. For the former, Macmillan can coordinate teams across the company to collaboratively build solutions with the goal of increasing reusability.
By structuring the AI approval process this way, McMillan hopes to allow banks to innovate without sacrificing security.
“While there may be a creative tension between experimentation and process, I believe that a rigorous process ultimately allows technology to be developed and deployed more quickly and efficiently. “There are,” McMillan said.
Inside the 8-step process
Anyone in the company can participate in proposing an AI solution, but it requires some effort. Primarily, employees must complete specific training on governance and AI principles and meet standards regarding information security.
The AI Steering Group meets biweekly to listen to feedback and typically considers five to six proposals. The steering group typically approves or approves with conditions, such as rethinking aspects of the solution or collaborating with other teams who have proposed similar ideas. In some cases, proposals are rejected, but McMillan said he usually tries not to do that.
“I don’t want to be in the position of saying ‘no’ to people. I want to say ‘yes’ to people and this is the best way to get to that,” McMillan said.
If your presentation is approved, the next steps typically include identifying who will deliver and understanding who needs to be involved in technology, legal, compliance, and risk. Employees running this process must also define deliverables, identify risks, and develop plans to mitigate those risks. This could be a standard set of questions and answers used for testing or to make certain teams aware of potential risks.
You also need to put together a business value proposition that outlines quantifiable benefits, such as reducing margins and operating costs, discovering new revenue streams, and reducing risk.
The AI Steering Committee meets biweekly to review the status of these projects. At the end of the process, the group pitch presents a final time to the steering group for approval to begin operations to ensure all conditions are met. Finally, the use case goes into production.
“What we do is we help them prioritize. We group them together and then my team hands you over. We say, ‘Okay, What are you trying to do?” and help set up the environment. Make sure you’re using the appropriate level of API. We are here for you as you work through your legal, compliance and risk processes,” McMillan said of his business partners.
