The Den Reviewthe briefing for executives who run AI
Wednesday, November 11, 2026

The Denominator

Subscribe
Trending

AI Wealth · AI Advisers

Morgan Stanley Advisors Reportedly Made Nearly 50,000 Attempts to Build Their Own AI Agents

One agent flags daily outflows above a set amount. Next comes FA Assist, which drafts proposals.

· 2 min read

A long, dim executive floor of desks, armchairs and table lamps overlooks a city at dusk, with two people working at the far windows.
Photograph: AI-generated

Morgan Stanley lets advisors build their own agents, from a daily large-outflow screen to portfolio watchers, and is building FA Assist to call agents for them, according to a published report.

Most of those agents never get past early testing, so the count shows how many advisors tried, not how many stuck. Morgan Stanley reviews an agent after an advisor’s team adopts it, and the firm did not say how many reach that point.

What are advisors building?

Advisors have made nearly 50,000 attempts, using plain-language prompts. Koren Maranca, head of AI for the wealth management division, said some agents track the holdings their clients care about. Others scan an advisor’s whole book, pull every outgoing flow above a set amount and deliver the list each day.

Who reviews them?

Most never last. “Sometimes people build an agent, they use it twice, and then they never touch it again,” Maranca said. The firm does not review those orphans, and reviewing all 50,000 would be impossible.

An agent that an advisor’s team adopts is reviewed for accuracy and completeness after use. One that looks fit for the whole firm gets more extensive testing.

What is the firm building itself?

AI@MS, the firm’s own program, gives every advisor a set of agents. Among the most used are one that answers planning questions that once went to an internal call center and one that organizes the advisor’s day. Roughly half of advisor questions now go first to an AI agent, and more than 1,000 firm experts reviewed its outputs to define how complete an answer must be before release.

FA Assist, still in development, will pull in agents automatically from a conversation. An advisor will be able to analyze client portfolios, simulate recession scenarios, find candidate models and turn the result into a proposal. The goal is to remove “a lot of the point and click.”

What do advisors want from AI?

Schwab’s 2026 Independent Advisor Outlook Study of 1,235 independent advisors found 81% look to AI for research and analysis and 79% for document drafting.

Which review would a firm’s own advisor-built agents reach: none, a team’s or the whole firm’s?

NVIDIAMicrosoftOpenAIAnthropicGoogleAWSJPMorganCapital OneAT&TCharterComcastT‑MobileUnitedHealthCVS HealthTravelersWalmartSalesforceDatabricksSnowflakeCoreWeaveStripeAccenture