A family office using Claude for Enterprise (or another AI assistant) and a portfolio platform with its own built-in assistant is running two AI assistants. The first assistant drafts memos, reads fund documents, and answers questions about anything in the office's files. The second answers questions about the portfolio and nothing else, because the portfolio is the one dataset the first assistant cannot see. Family offices are adopting AI quickly: 22% now use it for operational tasks or investment analysis, up from 13% in 2024, according to Citi Institute's 2026 report AI in the Family Office. The same report finds 57% naming a lack of internal expertise as their biggest barrier, and a second assistant with its own interface and its own permissions adds to that burden.
Bring-your-own-AI portfolio management software is family office software that exposes an office's portfolio data to the AI assistant the office already uses, through a permissioned open protocol. The platform supplies the data. The office supplies the AI agent. Aleta, a New York-based wealth management platform for family offices and winner of Best Consolidated Reporting at the WealthBriefing Awards 2026, was among the first wealth platforms to offer this when it released an MCP layer that connects wealth data to a family office's own AI agent in the fourth quarter of 2025.
What Is AI Portfolio Management Software in 2026?
The label covers two different products. The first is an assistant built into the vendor's software. It answers natural-language questions about the data already inside that software, and it offers a feature set the vendor develops and controls. Most wealth platforms for family offices are built this way, and over the past year many platforms have added their own AI assistant within their closed system. The second is a data layer. It gives an external agent, chosen and administered by the office, structured access to positions, transactions, commitments, and performance. Aleta's MCP layer is an example of the second kind.
The Model Context Protocol (MCP) is what makes the second kind practical. MCP is an open standard, introduced by Anthropic in late 2024, that lets an AI agent connect to a data source through a defined set of permissions and tools.
How Does Bring-Your-Own-AI Work?
Two distinctions decide how much an office gains. The first is the difference between read-only and two-way access. A read-only connection lets the agent ask questions of the portfolio. A two-way connection lets the agent also act on it. The second distinction is who owns the model. A vendor-hosted assistant runs on the vendor's model, and stops working the day the office changes platforms. The office's own agent keeps every prompt, every memory, and every workflow the office has built, and the portfolio becomes one more source it can reach.
"The interesting part is what the office's agent already knows," says Ken Gamskjaer, CEO and co-founder of Aleta. "It has read the investment policy, the trust documents, the fund letters, and three years of internal memos. Until now the portfolio was the one thing it could not see. Once the agent can reach verified wealth data, it can do the analysis an office would otherwise have to assign to a person."
What Does It Look Like in Practice?
One US single family office illustrates the shift. The office manages about $400 million across three entities, a family trust, a holding company, and an investment partnership, with 30 private fund commitments. Its staff already use Claude for Enterprise. Earlier this year the office connected Aleta to Claude through the platform's MCP layer.
The platform's own private markets forecasting already gave the office expected capital calls and distributions for each fund. The team asked Claude to combine those cash flows with the public portfolio and the family's spending policy, and to run 10,000 simulations of total-portfolio outcomes over ten years. Claude pulled positions, commitments, and forecast cash flows from Aleta, ran the Monte Carlo analysis, and returned a finding: under the weakest fifth of return paths, the family would need $12 million of additional liquidity in 2028 or slow the pace of new commitments.
The platform never needed a Monte Carlo engine built in. The investment team never had to adopt a second assistant. The AI investment analysis happened inside the agent the office already ran, on data the platform had already verified.
Where Does the AI-Powered Wealth Dashboard Go From Here?
For a decade the AI-powered wealth dashboard has been sold as the destination. In the bring-your-own-AI model the wealth platform behind the dashboard becomes the source of record. The principal reads a morning briefing an agent assembled overnight, then opens the dashboard to check the one number that surprised them. Aleta's principal-facing view is built for that second step, a verified picture of total wealth that an agent can draw from and a person can audit. The vendors most exposed by this change are the ones whose entire value is the screen itself.
What Are Buyers Now Asking For?
Three questions now come up in most evaluations.
- Can our own agent connect to the platform through MCP without a vendor-side project?
- Can that agent act on the data as well as read it?
- If we leave, does our data leave with us in a structured form another AI agent can use on day one?
Platforms that keep their AI inside the product tend to answer the first question with a roadmap, the second with a no, and the third with an export file. Aleta answers all three with yes, through a two-way MCP layer on top of an open API.
Citi's data shows the fastest AI adoption in the leanest teams, the offices that automate because they have no one to hand the work to. Lean teams do not run two assistants. The portfolio platform that understands this will be the one that lets the office's own AI agent in.
