Morgan Stanley Advisors Have Built 50,000 AI Agents
Morgan Stanley wealth advisors have created nearly 50,000 AI-agent experiments with natural-language prompts. Most are abandoned, while widely used tools undergo accuracy and completeness reviews.
Morgan Stanley allows wealth advisors to create artificial intelligence agents for routine tasks, producing nearly 50,000 experiments through natural-language prompts. Most are abandoned after limited use. Agents adopted by advisor teams undergo reviews for accuracy and completeness.
The program is intended to reduce routine work and give advisors more time with clients. Advisors have created agents to track portfolio holdings relevant to individual clients and to identify daily outgoing transfers above a set amount.
Koren Maranca, head of artificial intelligence for Morgan Stanley’s wealth management division, described the program this way: “If you want to build your own agent, and if you have a level of knowledge and complexity, the world is your oyster.”
The number of experiments does not represent the number of agents in regular use. Some advisors test an agent once or twice and stop using it. Morgan Stanley does not conduct a full review of these inactive tools.
Agents that become useful to several people on an advisor’s team receive a review of their accuracy and completeness. Tools being considered for use across the firm undergo more extensive testing. More than 1,000 Morgan Stanley employees from across the company have reviewed model responses and helped define standards for complete answers.
The firm also provides a shared set of AI tools through its AI@MS platform. The tools answer financial-planning questions, organize advisors’ schedules and estimate potential income from money placed in Morgan Stanley’s Signature Savings accounts.
About half of advisor inquiries are initially handled by an AI agent, according to Maranca. An advisor asking how to close a trust account with a remaining balance, for example, may receive an answer from the system before contacting an internal service desk. When the AI cannot answer a complicated question, it directs the advisor to the appropriate support team.
Another widely used tool prepares a daily morning briefing for advisors who opt in. The briefing can include upcoming appointments, suggested priorities and background material for client meetings. It may run to several pages and include information and recommendations related to a scheduled meeting.
Morgan Stanley has not assigned a specific number of hours that AI saves advisors each week. Company executives have said they expect the technology to help existing advisors serve more clients while leaving financial-planning decisions that require judgment and experience to people.
The firm began working with OpenAI’s technology within months of ChatGPT’s release in 2022. Its internal systems have since gone through at least two rounds of development. Morgan Stanley also blocks outside bots from accessing its websites and client accounts, in line with restrictions at some other large financial firms.
Morgan Stanley is developing AI@MS FA Assist, which is intended to reduce the need for advisors to select tools manually from an online dashboard. The planned system would let an advisor use a natural-language request to analyze portfolios, simulate risks such as a recession, identify investment models and assemble the results into a proposal. The feature remains in development and has not replaced the existing agent-selection process.








