Morgan Stanley will let corporate clients connect their AI agents directly to ShareWorks and Equity Edge, the platforms that administer stock plans for nearly half the S&P 500 and eight of the ten largest unicorn startups. A handful of clients already have access. The remaining 3,400 are scheduled for next year.
The connection runs on the Model Context Protocol, the open standard that links AI systems to enterprise software. Mark Mitchell, chief product officer of Morgan Stanley at Work, described the endpoint: clients won't log into ShareWorks or Equity Edge at all. Their interactions will be "purely agentic."
The AI agent replaces the software interface. The advisor stays.
The Funnel
Morgan Stanley built this channel through two acquisitions: Solium Capital in 2019, which became ShareWorks, and E-Trade in 2020. Together they created a pipeline where employees at 3,400 companies manage equity compensation through Morgan Stanley platforms. That pipeline holds $1.2 trillion in assets and feeds directly into the firm's $7.35 trillion wealth management division.
The conversion works like this. An employee's RSUs vest. Their equity grows. Morgan Stanley surfaces the pitch: let us manage the rest of your portfolio. The workplace channel isn't the product. It's the top of the funnel.
Opening that funnel to AI agents means corporate finance teams can pull vesting data, run compliance reports, and query employee equity positions through software instead of a web interface. The work is repetitive, high-volume, routine. Stock grants follow templates. Vesting schedules run on calendars. Option exercises are arithmetic. This is exactly the category where agents already outperform humans: retrieving structured data from systems designed for human eyes but that never needed human judgment.
Morgan Stanley doesn't need thousands of new employees to scale the funnel. It needs agents to process the queries that already flow through it.
The Standard
The choice of MCP over a proprietary API is the decision that makes this worth writing about.
A proprietary API would lock clients into Morgan Stanley's tooling. MCP lets any agent connect to the same platforms through the same protocol. Morgan Stanley is betting that the value lives in the proprietary data and business logic behind the connection, not in controlling the connection itself. Mitchell said as much: "The companies that are going to survive in the future are the ones who have proprietary data and business logic, which is the foundation of our offering."
This inverts the walled-garden strategy that dominated fintech for a decade. Morgan Stanley is opening the interface layer because its moat sits underneath it. The protocol is generic. The data isn't.
Robinhood made a similar MCP bet in May when it launched agent trading accounts. But Robinhood opened MCP for retail execution: agents placing trades, managing portfolios, spending through virtual credit cards. Morgan Stanley opened MCP for institutional plumbing: agents pulling equity compensation data for corporate finance teams. Same protocol. Different layer of the financial system.
The Gap
JPMorgan Chase and Goldman Sachs both deploy AI agents internally. Code generation. Document summarization. Operational acceleration. Neither has announced plans to let external agents connect to client-facing systems. All three banks have the engineering talent to build this. What separates Morgan Stanley is willingness to accept a specific kind of exposure: letting someone else's software touch its platform through an open interface, in a regulated industry where every data request creates liability.
Morgan Stanley can move first because the workplace channel is the right place to take this risk. Stock plan administration is standardized enough that agent access doesn't threaten the judgment layer. It accelerates the plumbing that feeds the advisory relationship. And it offers a natural experiment: if something goes wrong with an agent pulling vesting data, the blast radius is contained. No trades executed. No money moved. Just data retrieved.
If JPMorgan or Goldman opens external agent access within six months, the first-mover advantage is modest. If they don't, Morgan Stanley will have spent a year building agent integrations with half the S&P 500.
The entry point matters. The first major bank to open its systems to AI agents didn't open advisory. Didn't open trading. It opened the filing cabinet where equity compensation records sit: $1.2 trillion in assets flowing into a $7.35 trillion wealth management operation.
The boring work came first. In financial services, it always does.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)