Family office research
AI for family office manager due diligence: an evidence checklist
A complete-looking summary may still leave its evidence unclear. Family office research can begin with a common format that takes each claim back to a source or calculation. This FiClaw team checklist is an original workflow design, not a manager assessment, allocation recommendation or standardized due diligence questionnaire.
Published by FiClaw 产品团队 · Published · Updated
阅读中文版Quick answer
Start AI-assisted manager due diligence by separating manager statements, independent checks, research experiments and open questions. Record the source, date, location, reviewer and next action for each entry. FiClaw currently supports explicit A-share quant research; document repositories, a due diligence workspace and committee preparation workflows need pilot assessment. An AI summary does not complete due diligence.
Key takeaways
- Label manager-provided statements separately from independently checked facts.
- An independent strategy experiment cannot authenticate a manager's live performance.
- Hand off interview questions and evidence gaps alongside human conclusions.
Replace unsupported summaries with an evidence matrix
For each material claim, record the original content, document version and date, location, verification status, reviewer and next step. Retain earlier versions and explain changes. When the source cannot be located, mark the claim unverified.
AIMA's 2025 manager DDQ update discusses performance presentation, technology and outsourcing, liquidity and fees among its themes. It can inform interview coverage. The table below is our own research record format, does not reproduce the questionnaire, and implies no AIMA endorsement or certification of FiClaw.
| Category | Example record | Next action |
|---|---|---|
| Manager statement | A document describes drawdown controls; retain its page and date | Ask about conditions, exceptions and supporting evidence |
| Independent check | Identify the independent source actually obtained and its scope | Record reviewer, date and uncovered questions |
| Research experiment | Test explicit rules; retain code and configuration | Explain differences from the manager's actual strategy and execution |
| Open question | Necessary fee, liquidity or information-timing detail is absent | Assign an owner, interview question and follow-up date |
Turn a statement into an answerable interview question
Turn 'strict risk control' into questions about the constraint, trigger, exception authority and checkable records. Break a capacity claim into the capital scale, execution assumptions, concentration and stressed conditions that need explanation. Researchers should select questions appropriate to the manager and available materials.
Do not ask AI to invent complete answers for missing documents. Confirm the source and scope of screenshots, second-hand accounts and unclear versions before drawing conclusions. Automated bulk ingestion and cross-document citations should be validated as separate product pilot requirements.
Research object and question: [manager / strategy and scope] Permitted material: [authorized sources, versions, dates and redaction] Organize these fields without inventing facts: 1. Original statement and source: [document, page or paragraph] 2. Category: [manager statement / independent check / experiment / open question] 3. Confirmed scope: [evidence and limits] 4. Follow-up: [question, requested material and owner] 5. Conflicts or missing evidence: [keep unresolved] 6. Human review: [completed by the researcher] Use for interview preparation, without automatic ratings, allocation percentages or investment conclusions.
What a FiClaw research experiment can establish
Where material describes an implementable A-share hypothesis, use FiClaw to organize a specification, generate Python and run historical tests and diagnosis. Confirm data and interface support first; disclose missing historical constituents, information timing or execution conditions.
The experiment tests your defined rules. Even a similar-looking curve does not establish replication of the manager's account, fees, execution, leverage or live record. Do not present independent research as performance authentication without adequate evidence.
Prepare committee research materials with visible gaps
- 01Fact list: source, date, verification status and reviewer for each conclusion.
- 02Experiments: hypotheses, data scope, configuration, results and unsupported conditions.
- 03Questions and disagreements: outstanding materials, conflicting statements and follow-up interviews.
- 04Human interpretation: how the team reads the evidence and which judgments still need professionals.
- 05Scope: research discussion material does not replace manager due diligence, legal or tax work, or final allocation decisions.
Frequently asked questions
Can a complete AI summary conclude manager due diligence?
No. A summary supports reading but cannot authenticate material or replace interviews, independent checks and professional judgment. The institution's existing diligence and approval processes determine completion.
Can a family office reproduce a manager's live record with FiClaw?
Do not assume so. FiClaw can test explicit rules within supported data scope. Actual accounts, executions, fees, leverage and strategy details may be unavailable, so an independent backtest cannot authenticate a manager's performance.
Is a complete family office due diligence system already available?
FiClaw's current public core is a quant research workstation. Bulk material ingestion, cross-document evidence citations, a manager diligence workspace and committee collaboration require separate scoping and assessment.
Method and product references
Related reading
Discuss one family office research task
Use public or redacted samples to explain evidence and review needs, then agree a suitable pilot scope.