The silent check
Private wealth is won by introduction. Families now verify the introduction with AI, before the first meeting.
The problem: word of mouth now has a second opinion
In Swiss private wealth, new mandates rarely come from advertising. They come from a lawyer, a banker or another family who knows the partners. That has not changed.
What has changed is the evening after the introduction. The family, often the next generation, asks ChatGPT, Gemini or Claude about the house. The engine gives one confident answer, not ten links.
If that answer leaves the house out, quotes the wrong fees or mixes it up with another company, doubt sets in quietly. The first meeting is postponed, or never happens. Nobody tells the house why.
We call this the silent check. It is invisible to every tool a house uses today, because it happens in a private conversation between a family and a machine.
The house
House M is an independent multi-family office in Geneva, founded in 1998. Three partners advise about 60 families with CHF 2.4 billion in assets. Nearly every new mandate comes through an introduction.
Over two years, the partners saw introductions convert less often: from about one in three to one in five. The families were the same kind. The partners were the same. Something between the introduction and the meeting had changed.
The first reading
We asked the five core questions, plus six of the house's own, on four engines. We asked in French, English and German, as families in Geneva, Zurich and London would. 240 answers were sealed in one month.
Result: a Narrative Visibility Score of 0.27, tier Critical.
- Q1 LeadershipAbsent in 92% of runs
Asked for the best independent houses, engines named six competitors and never House M.
- Q2 FeesDiscounted in 64% of runs
Engines quoted fees of about 1.2% a year. House M charges 0.55%.
- Q3 ProvenanceNamed in 70% of runs
The founding year and partners were mostly right, but the regulator was often wrong.
- Q4 SubstitutionSubstituted in 71% of runs
Asked for alternatives, engines steered the family to two larger Zurich houses.
- Q5 Due diligenceConfused in 38% of runs
Engines mixed House M with a foreign property developer of the same name, and its lawsuits.
The engines knew House M existed. They simply never volunteered it, and when asked directly they described it with someone else's fees and someone else's lawsuits.
What we did
House M took a six-month Mandate. Agents prepared every action; one partner approved each one before it went out. Nothing was published in the house's name without that approval.
- Month 1
Correct the record
We traced each wrong answer to its sources: two outdated directories, an old press article and a registry entry. Each was corrected.
- Month 1
State the facts plainly
The house site gained one clear page of facts: ownership, regulator, fee model, team and history, written so engines can quote it.
- Month 2
Separate the namesake
We anchored the house identity in the registries and profiles engines trust, with its full legal name, address and regulator.
- Months 2–4
Earn the right citations
We identified four publications engines cite for Swiss family offices. Agents prepared two expert pieces and one profile for the partners to approve.
- Months 3–6
Measure and adjust
Each monthly reading showed which actions moved the answers. Actions that did not move anything were dropped.
The readings, month by month
Every month the same questions were asked again, on the same engines, in the same markets.
| Measure | Month 0 | Month 2 | Month 4 | Month 6 |
|---|---|---|---|---|
| Narrative Visibility Score | 0.27 | 0.38 | 0.49 | 0.61 |
| Tier | Critical | At-risk | Monitor | Monitor |
| Q1 absent | 92% | 81% | 63% | 48% |
| Q2 fees misstated | 64% | 30% | 18% | 12% |
| Q4 substituted | 71% | 66% | 52% | 39% |
| Q5 confused with namesake | 38% | 14% | 7% | 4% |
Fees and identity were fixed first, because they depend on facts the house controls. Leadership and substitution moved more slowly, because they depend on what others write about the house.
What it meant for the house
In the six months after the Mandate began, introductions converted to a first meeting about one time in three again. Two new families mentioned that they had checked the house with AI before calling.
The partners now open each monthly reading before their partner meeting. They see which answers changed, and which competitor gained ground on which question.
Lessons for any house
Introductions still win
AI does not replace the introduction. It decides whether the introduction survives the first evening of research.
Silence is not safety
A house that says little in public leaves the engines to fill the gap with competitors, namesakes and guesses.
Facts beat promotion
What moved the answers was accurate, quotable facts in trusted places, not more marketing.
Discretion and clarity fit together
House M published no client names, no performance and no rankings. It only made its own facts clear.
A private reading of your house
We work with a small number of houses each season, under a confidentiality agreement. Most come through an introduction. You can also ask us directly, below.