The office that runs on discretion.

A family office is an information-control institution before it is an investment institution. It exists because certain statements, structures, and correspondence must not circulate — not through an advisor’s CRM, not through a vendor’s analytics, not through anyone’s training data. The staff is small on purpose. The perimeter is the product.

The family office already knows AI well — as an investment. UBS surveyed 307 offices this spring and found AI the leading theme in their portfolios, with 65% invested somewhere across the value chain. Inside the office itself the technology arrived the way consumer technology always arrives: informally. The latest RBC and Campden Wealth study of North American offices found three times as many using AI in their operations as the year before. A browser tab here, a personal subscription there, an assistant pasting a capital statement into a chat window before a call. So the office is engaged with AI twice over — as a conviction, governed by an investment committee, and as a habit, governed mostly by nobody.

It is tempting to file this under technology adoption. That misses what is actually at stake. The adoption is not the risk; the informality is. An institution built to control channels has acquired an unexamined one, and the acquisition happened without a decision — which is the precise kind of event a family office exists to prevent. The right analogy is not a firm buying software. It is a household taking on staff without references.

The standard fixes fit badly here. Enterprise governance assumes headcount the office does not have and does not want. A technology hire is hard to justify for a team of six and sits awkwardly in one. And prohibition, the other reflex, mostly converts visible informality into invisible informality — the tools are useful, and the people using them know it.

What fits is the instrument the office already trusts: a defined mandate. Decide what is sanctioned and on what terms. Sort, even roughly, what may enter which systems — for many offices the first time anyone has written that down. Look hard at the few workflows where something purpose-built would earn its keep, and be willing to conclude, for most, that the right answer is nothing. Build the few that clear the bar inside an environment the office controls. Put a name on it. This is not enterprise IT. It is the same structure the office applies to custody, to trustees, to anyone else who touches the family’s affairs.

What makes the exercise worth its cost is what sits on the other side. The characteristic burden of a small office is fragmentation: decades of statements, K-1s, trust documents, minutes, and letters spread across systems, inboxes, and in some cases filing cabinets, held together by the memory of two or three long-tenured people. Memory that lives in people retires with them. A system that brings the office’s own record together, privately, and can answer questions about it is not a convenience. It is the lean office’s oldest ambition — more capacity without more headcount — applied to the asset it has always been richest in and least able to use.

A limit, plainly stated. Much of what a family office does should never be systematized: the judgment, the relationships, the sense of when a beneficiary needs a call instead of a report. Discretion includes knowing what not to automate. The offices that get this right will not be the ones that adopted the most. They will be the ones where every channel, old and new, was chosen.