Use Cases

AI agents at work

Four pieces of real work with family offices and the teams around them. We keep client names private, so what you read here is the workflow and the result.

Independent Equity Research · Switzerland

Quarterly model review, from 120 hours to 12

120h → 12h Quarterly model review, per analyst

The situation

Every quarter, an analyst spent around 120 hours updating the financial models for the listed companies the firm covers: pulling figures out of company reports, validating them, and cross-referencing them. Slow, manual, and exposed to human error at exactly the points that matter most.

What we deployed

A Due Diligence Agent, running inside the firm's own environment. It extracts the numbers from the company reports, checks them against the models, cross-references across documents, and flags anything that does not reconcile for a person to review.

The result

The quarterly review dropped from 120 hours per analyst to 12, delivered inside one week. Along the way the agent surfaced five forecasting errors that had gone undetected across several analysts and several years. The analyst's time moved from data gathering to judgment.

Family Office · Private Equity

From the memorandum to the valuation model

1 memorandum → 3 documents Acquisition screening, from the document into the model Watch it run

The situation

A private-equity team screens dozens of acquisition memoranda a year, each one eighty to a hundred pages. Reading it, pulling the figures out, and typing them into a valuation model costs one to two days of senior time before anyone can judge the deal.

What we built

Two agents in sequence. The first puts the same fifteen due-diligence questions to every document, and every figure it returns opens on the page it came from. The second writes those figures into the team's own Excel model. The agent fills the inputs, the arithmetic stays in the spreadsheet the team already trusts.

What it produces

Run on published memoranda, it returns a summary of the deal, the model with its inputs filled in, and a draft non-binding offer. Of 462 figures tested, 456 carried a link to their source page, and the six that did not were flagged for a person to check.

Investment Holding · Private Equity Portfolio

Portfolio monitoring across sixty-plus holdings, on a recurring cadence

60+ Portfolio holdings monitored, every report fact-checked before delivery

The situation

A multi-entity investment holding produces a standardized monitoring report for each company in its private-equity portfolio. Built by hand, the reporting could not keep pace as the portfolio grew, and consistency slipped from one report to the next.

What we deployed

A reporting agent inside the holding's own Microsoft Azure environment. It reads the portfolio data where it already sits, generates one standardized report per holding on a recurring cadence, and runs every output through a battery of fact-checking tests on names, numbers, and dates before it reaches the team.

The result

The engine produces the full report set across a portfolio of more than sixty holdings, each report validated before delivery. The same quality holds whether it is three companies or sixty, and the reporting stays current as the portfolio expands.

High-Net-Worth Advisory

A full workflow diagnosis, before a single agent is built

Diagnosis A sequenced build plan before a single agent goes live

The situation

A founder-led advisory serving high-net-worth private clients wanted AI across the practice, but first needed an honest read on what agentic AI can and cannot do for each part of the work, and in what order to build.

What we did

We ran a structured diagnosis of the practice: its workflows, documents, and tools, mapped against where an agent adds real value. The output is a findings report and a build sequence, with the prerequisites and the honest limits named up front.

What it surfaced

Clear early priorities emerged: a daily briefing that consolidates the principal's tasks across inbox, messages, and notes; turning one flagship market report into many pieces across channels; and a searchable knowledge base for templates and past work. The principal has a sequenced plan in hand before committing to a single build.

Reporting our team assembled by hand every cycle now comes out drafted by an AI agent, inside our own environment. A positive outcome.
A family office in Switzerland

Engagements are presented anonymously. Named references are shared privately, with client permission, on request.

AI agents at work

Quarterly model review, from 120 hours to 12

The situation

What we deployed

The result

From the memorandum to the valuation model

The situation

What we built

What it produces

Portfolio monitoring across sixty-plus holdings, on a recurring cadence

The situation

What we deployed

The result

A full workflow diagnosis, before a single agent is built

The situation

What we did

What it surfaced

See it working on your data.