Family Office InsightsMaria Skotheim

Financial Ontology for Family Offices: Why AI Needs More Than Account Balances

A financial ontology is the map of how a family's wealth fits together: who owns each asset, which entity holds it, how entities and people relate to one another, and which documents support each investment. A semantic layer makes sure the numbers drawn from that map, such as performance, liquidity and exposure, are defined and calculated the same way every time.

Both matter more for a family office than for most investors, because family wealth rarely sits in one place. And both matter more once AI is involved, because AI can only answer questions about relationships and definitions the system already understands.

Why account balances aren't enough

Consider a family with a fairly common structure:

  • A family trust owns 100% of a holding company.
  • The holding company owns a commercial property and an interest in a private equity fund with $2 million still uncalled.
  • Two adult children are beneficiaries of the trust, and one also holds a personal brokerage account.
  • The property has a loan secured against it, held at the holding company level.

A custodial feed shows the brokerage balance and the holding company's cash account. A spreadsheet tracks the property value and the loan. The fund's latest capital account statement and capital call notice sit in someone's inbox.

Each figure may be accurate on its own, but none of them explain how the pieces connect. How much is the trust worth? How much private equity is held in the trust for the beneficiaries? Does the holding company have enough cash for the next capital call after the loan payment? The answers depend on relationships that don't exist in any single source.

What a financial ontology does for a family office

An ontology turns disconnected records into a structure the system understands. In practical terms, it captures:

  • Ownership: which person or entity owns each asset, and in what proportion.
  • Entity relationships: how trusts, holding companies, partnerships, foundations and operating businesses own, or are owned by, one another.
  • Stakeholder roles: who the beneficiaries, trustees, settlors and advisors are, and how each relates to each entity.
  • Supporting documents: which subscription agreements, capital call notices, K-1s, loan agreements and trust documents belong to which investment or entity.
  • Obligations: unfunded commitments, loans and other liabilities, and which entity is responsible for each.

With that structure in place, the trust's value rolls up through the holding company: the property net of its loan, the fund interest and the holding company's cash. A capital call notice is linked to the fund, the holding company that holds the interest and the trust above it. Reporting for each beneficiary can show what they own directly and what is held for them through the trust.

The same structure supports day-to-day work. When a distribution notice arrives, it can be recorded against the right fund and entity, and the family's reporting reflects it without anyone re-entering the figure in a separate spreadsheet.

What a semantic layer adds

The ontology tells the system how things connect. The semantic layer tells it what the numbers mean.

Family offices regularly receive figures calculated in different ways. A fund administrator reports one measure of performance, a custodian another and an internal spreadsheet a third. Available liquidity might include a credit line in one report and exclude it in another. Unfunded commitments might be current in one place and a quarter out of date in another.

A semantic layer sets one definition for each measure, such as how IRR is calculated, what counts as liquid and how exposure is attributed through entities, and applies it consistently across every report and every question.

Where data reconciliation fits

The ontology and the semantic layer both assume the underlying records are right. Reconciliation is the work that checks: matching balances, positions and transactions across custodians, fund administrators, banks and internal records, and flagging breaks, duplicates, gaps and stale data.

In the example above, reconciliation is what catches a capital call paid from the holding company's account that hasn't yet appeared on the fund's statement.

Put simply, the ontology says how things relate, the semantic layer says what the numbers mean, and reconciliation confirms the numbers are complete and current.

Why AI needs both

Suppose a family principal asks an AI assistant:

How much cash will the trust need for capital calls over the next 12 months, and can it cover them without selling anything?

To answer accurately, the AI needs four things:

  • From the ontology: which fund commitments roll up to the trust through the holding company.
  • From linked documents: what is still unfunded and what the latest capital call notices say.
  • From the semantic layer: what counts as available liquidity, and whether cash set aside for the loan payment is excluded.
  • From reconciliation: whether the cash balances are current and complete.

If any of those pieces is missing, the AI fills the gap with assumptions. The answer can sound complete while missing a commitment held through a different entity or counting cash that is already set aside for a loan payment.

AI is only as useful as the data and workflows behind it.

Building the foundation vs. building the AI

Family offices can and do build their own AI tools. The AI interface is the more accessible part. The ontology, the definitions and the workflows that keep them accurate as entities are formed, investments change and new documents arrive are the larger, ongoing project.

How MYFO provides the foundation

MYFO is an AI-native family office operating system built on a financial data ontology. It maps ownership across individuals, trusts, holding companies and partnerships, captures stakeholder roles such as beneficiaries and settlors, and links investments to their supporting documents.

Established workflows handle ingestion, parsing and reconciliation, including private investment documents such as capital call notices, distribution notices, K-1s and fund statements. Performance calculations are applied consistently within the platform, so the same figures appear in reports, dashboards and AI-generated answers.

MYFO works alongside a family's existing general ledger and can be used with or without AI. The same structured data supports MYFO's own reporting and AI capabilities, Claude through MYFO's MCP integration, and other tools and models the family chooses to use. That gives AI a connected picture of the family's wealth to work from, instead of account balances and disconnected files.

For a broader look at these concepts across financial institutions, read Financial Ontology vs. Semantic Layers Explained.