GenAI does not just make investment teams faster. It changes the economics of bespoke service, allowing firms to extend more individual attention beyond their largest clients.
In June 2026, CNBC reported that Morgan Stanley was opening Shareworks and Equity Edge, its workplace stock plan platforms, to AI agents used by corporate clients. The change does not apply to Morgan Stanley’s entire wealth management business. It is still an important shift. Instead of requiring every user to navigate an interface designed for people, the platforms can allow those agents to retrieve relevant data and insights on a client’s behalf.
That distinction matters. A portal gives many clients the same interface and leaves each person to locate what is relevant. An agent can begin with the user’s context and assemble a more specific response. The software performs more of the retrieval and initial tailoring that previously required manual work.
Morgan Stanley’s internal experience shows that this model can move beyond pilots. According to an OpenAI case study, more than 98% of Morgan Stanley adviser teams use AI @ Morgan Stanley Assistant. The share of documents accessible through the system rose from 20% to 80%. A separate meeting tool, AI @ Morgan Stanley Debrief, turns consented Zoom recordings into draft notes and follow up messages, which advisers review before finalising.
The point is not that the human disappears. It is that the machine can produce the first tailored output, while the human focuses on judgment, approval, and the client relationship.
Bespoke Was Rationed by Human Effort
Historically, every tailored portfolio, investor letter, and considered response consumed someone’s time. The cost increased with every client and every request, so firms rationed bespoke service. Senior attention went to accounts where it produced the greatest commercial return. Other clients received more standardised products and communications.
GenAI changes part of that arithmetic. It does not make personalisation free. Firms still pay for data integration, model usage, security, evaluation, and review. But once those foundations are in place, generating another draft, summary, or explanation tailored to a client can cost far less than producing it entirely by hand.
That is the real economic shift. Customisation can move from a service rebuilt for every client to a repeatable capability governed by common data, rules, and controls.
The wealth channel offers a useful precedent. BNY notes that direct indexing was historically concentrated among wealthy investors because customised portfolios required significant cost and human effort. Technology has helped lower operating costs and widen access. GenAI can extend the same logic from portfolio construction into reporting and communication.
Three Layers of Personalisation in Private Markets
In private markets, personalisation can develop across three layers: how information is presented, how it is explained, and what the client owns.
The first is reporting. A workflow supported by AI can create versions of a master report for individual LPs using documented preferences for format, currency, language, or attribution method. V7 describes this capability in its LP reporting platform, but this is an example reported by a vendor rather than evidence of adoption across the industry. The credible use in the near term is not autonomous reporting. It is automated assembly followed by reconciliation, compliance review, and approval.
The second is communication. A model connected to approved records can draft an update that reflects an LP’s mandate, prior questions, and relevant portfolio exposures. Morgan Stanley’s Debrief tool provides a useful adjacent example: it generates CRM notes and follow up drafts, but advisers review and adjust the outputs before sending them. In a private markets setting, the same pattern could turn a generic quarterly message into a more relevant first draft while keeping authority over facts and commitments with the firm.
The third is the mandate itself. Blackstone says its insurance business offers solutions ranging from customised separately managed accounts to full portfolio management, with products aligned to each insurer’s objectives, risk, and outlook. These mandates predate GenAI. The opportunity for AI is operational: helping teams monitor, report on, and communicate across more bespoke mandates without increasing manual work at the same rate.
This distinction keeps the argument honest. GenAI is not creating demand for custom portfolios. It is lowering some of the operating friction involved in serving them.
Why This Is Happening Now
The operating challenge is already large. On its current private wealth webpage, Blackstone reports $324 billion in private wealth assets under management and 357 dedicated professionals. BNY also markets infrastructure designed to rebalance tens of thousands of customised accounts in less than a trading day. These investments indicate that providers already view personalisation at scale as a strategic and operational priority, even before GenAI is added.
The commercial inference is straightforward. As firms serve more individual accounts and offer more customised structures, human capacity becomes the constraint. A firm cannot increase senior attention in direct proportion to account growth.
The practical answer is not limitless automation. It is a division of labour. Software handles retrieval, transformation, and first drafts. People retain responsibility for exceptions, judgment, and decisions with serious consequences.
A broader client base creates the demand. GenAI can make the service model more affordable.
Scale Requires Control
For investment firms, caution is necessary. The SEC states that investment advisers owe clients a fiduciary duty comprising duties of care and loyalty. It also makes clear that automated advisers remain subject to the Advisers Act. An output generated by AI does not sit outside those obligations simply because software produced it.
The technology brings its own risks. NIST identifies confabulation, data privacy, information integrity, and excessive reliance as important concerns associated with generative AI. These risks become more consequential when a common model or workflow produces outputs for many individual clients. A flawed rule, stale source, or unsupported statement can be repeated at scale.
Controls therefore have to be part of the service design. Morgan Stanley reports using evaluations before deployment, daily regression testing, expert feedback, and adviser review of generated outputs. For private markets firms, the equivalent should include approved source data, traceable citations, access controls, version history, testing against realistic LP scenarios, and human review proportionate to the consequence of the output.
This is where a careful alternatives firm may have an advantage. Disciplines for model risk, formal approvals, and audit trails are not obstacles to personalisation. They are the infrastructure that makes it deployable.
The next decade will not belong to the firm that advertises the most customisation options. It will belong to the firm that gives more clients relevant, bespoke attention without adding a person for every account and without weakening control.
These are the operating questions we will be working through with enterprise and investment leaders at GAI World 2026, taking place September 28 to 30 at the Hynes Convention Center in Boston.
GAI Insights Team
