Morgan Stanley Moves Wealth AI From Answers to Actions

Morgan Stanley's wealth business is moving artificial intelligence from answering questions to carrying out controlled workflows across multiple systems. That shift could save advisers and service staff time, but the bank has not yet quantified the revenue, cost savings or error rates tied to the new tools.

The strategy rests on work that began years before generative AI became popular. Hina Shamsi, chief information officer for wealth management technology, told CIO Dive that the division stopped organizing data only around individual accounts and began building a broader view of each client relationship.

That change supported machine-learning tools that recommend the next action for an adviser. It also created the context and semantic layer that newer AI agents need to understand data consistently across the wealth platform.

Three agents, different boundaries

Morgan Stanley is focusing on three areas: client interaction, service support and adviser assistance.

The client agent lets customers ask questions in plain language and complete common tasks. Management plans to add more account-specific context over time.

The service agent helps client-service associates retrieve information and complete work such as moving money, opening accounts and changing beneficiaries. These tasks sound simple, but they often require several applications and approval steps. Automating the navigation can reduce manual effort while leaving the professional responsible for the decision.

The adviser agent supports daily work so advisers can spend more time with clients. This builds on earlier products such as AI @ Morgan Stanley Debrief, which can create meeting notes and draft follow-up emails with a client's consent. The adviser still reviews the output before sending it.

Controls are part of the product

The expansion from search to action raises the stakes. An incorrect summary is inconvenient. An agent that moves money or changes a beneficiary can create financial and legal harm if permissions are wrong.

Morgan Stanley says its agents operate within defined limits covering the data they can access, the actions they may take and the points at which a human must intervene. The firm's public AI materials also emphasize human-centered design, data protection, monitoring and oversight.

The financial payoff is still unproven

The wealth unit has scale that can make small productivity gains meaningful. Third-quarter 2026 wealth-management revenue reached a record $9.07 billion, up from $8.23 billion a year earlier. The business reported $232.4 billion of net new assets and a 30.4% pretax margin.

Those results do not isolate the effect of AI. Morgan Stanley attributed the quarter mainly to higher asset-management fees, net interest income and client flows. Technology spending also increased. Without disclosed adoption, time saved, error rates or unit costs, investors cannot separate operational progress from a measurable earnings contribution.

The important change is architectural. Morgan Stanley is no longer treating AI as a stand-alone chatbot. It is connecting agents to workflows, permissions and client context. That could widen the bank's service capacity, but the next proof point should be evidence that the tools improve productivity without weakening control.