Morgan Stanley advisors have made nearly 50,000 attempts at building their own AI agents

Morgan Stanley lets its wealth management advisors build their own AI agents from natural-language prompts, and they have made nearly 50,000 attempts so far, Financial Planning reported on Nov. 2.1

The agents are designed in large part to give advisors more time with clients by automating routine tasks, according to the report.1 Koren Maranca, head of artificial intelligence for the firm's wealth management division, said agents are reviewed only after they gain use, because reviewing all of them would be impossible.1

Maranca said agents embraced by many people on an advisor's team get careful reviews for accuracy and completeness, and agents considered for firmwide adoption get more extensive testing.1

"Sometimes people build an agent, they use it twice, and then they never touch it again," Maranca said, adding that orphaned agents get no comprehensive review.1

Maranca said agents that become part of an advisor's practice go through review and approval post-use; most agents do not make it past early testing, according to the report.1

Examples of advisor-built agents include one that tracks portfolio holdings of interest to clients and one that pulls all outgoing flows above a certain amount across an advisor's book each day, Maranca said. The sources do not say what data access controls apply to these agents.1