> ## Content Index
> Fetch the complete content index at: http://localhost:2368/llms.txt
> Use this file to discover other available public pages before exploring further.

# A source-first research workflow for family office teams
- URL: http://localhost:2368/a-research-workflow-for-family-office-teams/
- Published: 2026-10-23T11:19:11.000Z
- Updated: 2026-10-24T12:21:19.000Z
- Description: Separate collection, interpretation and decision-making so that automation can help without obscuring the underlying evidence.
- Author: The Review Desk
- Tags: Family Offices, Technology, Insights

A research workflow becomes easier to review when it separates the source record from the interpretation built on top of it. That separation is particularly useful when AI assists with collection or drafting.

## Keep a stable observation

For each material figure, record the publishing organization, original URL, metric label, unit, measurement date and reporting scope. Keep the retrieval date separate from the date of the underlying observation.

## Use automation for bounded work

Collection, duplicate detection and basic format checks can operate against explicit rules. A model can help classify a document or draft an explanation using a supplied set of facts. A missing figure should not become an invitation to estimate one without disclosure.

## Make exceptions visible

A change in legal entity, a conflict between sources or a new performance claim needs attention. The workflow should retain the competing observations rather than overwrite history. Source changes are candidates for review, not automatic corrections to a published record.

## Connect the record to the article

The reader should be able to reach the relevant source from the article. The editor should be able to identify the observation used, the drafting process and subsequent changes. This makes a correction more useful than a silent replacement.

The local Review prototype follows this pattern for its founding data sample. It does not claim that model-generated text has been independently verified merely because another model agrees with it.

---

This article describes the Review's proposed editorial workflow. [Read the methodology](http://localhost:2368/methodology/).