What an AI research teammate should own
An AI research agent becomes a useful teammate when it owns a continuing research responsibility, not just a single search. For a family office or private-capital team, that job might be: prepare concise, sourced briefs on managers, companies, advisers, or co-investors before important conversations, then keep the open questions visible for the next review.
The teammate should not declare that a manager is investable or turn public information into an investment recommendation. Its value is narrower and more practical: assemble the relevant evidence, distinguish current facts from older claims, surface contradictions, and give the human reviewer a better starting point.
Meet your research teammate
| Responsibility | Prepare decision-useful, sourced briefs before manager and counterparty conversations. |
| Working style | Concise, evidence-led, dated, and explicit about unknowns. |
| Needs from you | The mandate, subject, purpose of the meeting, approved sources, and preferred brief format. |
| Brings back | A research brief, material changes, open questions, source links, and suggested conversation points. |
| Reports back | In its continuing conversation, with evidence cited and unresolved work visible. |
This is one role inside a broader AI teammate strategy for private capital. The same responsibility can support a family office researching a manager, a GP reviewing an LP, or a sponsor preparing for a capital partner. The research criteria must change with the mandate.
A useful brief is not a longer biography
Generic research often starts with a company description and ends with a list of news headlines. A private-capital brief needs a sharper question: what does the reviewer need to understand before the next decision or conversation?
For a family office meeting a venture manager, the useful structure may be:
- Identity and role: Who is the manager, which entity is involved, and who will attend?
- Mandate relevance: Which strategy, stage, geography, or exposure appears relevant to the office's stated criteria?
- Current evidence: What has the manager publicly disclosed, and when was each source published or updated?
- Relationship context: What prior meetings, introductions, requests, or commitments are available in the authorized workspace?
- Material changes: What changed since the last review?
- Open questions: Which important points remain unsupported, inconsistent, or unavailable?
- Conversation plan: Which questions would resolve the highest-value unknowns?
The ILPA Due Diligence Questionnaire illustrates the breadth of manager diligence. Its questions extend across organization, strategy, track record, operations, governance, service providers, legal matters, and requested documents. That makes it a useful research reference, but not a substitute for an investor's own diligence scope or professional advice.
Use an evidence hierarchy, not a pile of links
A research teammate should know what kind of evidence it is looking at. Use a hierarchy that keeps provenance visible:
| Evidence type | Example | How to use it |
|---|---|---|
| Primary public source | Firm site, regulatory filing, official announcement, current policy | Support a dated factual statement while noting the source's own perspective. |
| Authorized internal source | CRM record, meeting note, approved document, email thread | Preserve relationship-specific context and access boundaries. |
| Credible independent source | Reporting with named sources or transparent methodology | Add context or corroboration, not automatic truth. |
| Database or directory record | Licensed profile, public directory, adviser registration | Verify identity or reported fields, then confirm mandate-specific claims elsewhere. |
| Inference | A hypothesis derived from several signals | Label it as an inference and turn it into a question for review. |
For registered or reporting investment advisers in the United States, the SEC's Investment Adviser Public Disclosure service provides access to filed Form ADV information. It can help verify an entity and certain reported business or disciplinary information. A filing does not establish current investment appetite, fit, reputation, or the likelihood of a relationship.
Every material note should preserve three details: source, observation date, and confidence. Without them, a polished brief can conceal stale or unsupported information.
A reusable manager-brief format
Use this compact artifact to keep research tied to the upcoming conversation.
| Section | What belongs here | What does not |
|---|---|---|
| Why this matters now | Meeting purpose, active diligence question, or current relationship event | Generic importance or prestige |
| Verified profile | Current entity, team, strategy, geography, and reported activity | Assumed mandate based on one investment |
| Relevant evidence | Source-linked facts that bear on the review objective | Every available fact about the firm |
| What changed | New fund, role change, updated material, or corrected record | Undated headlines |
| Relationship context | Authorized history, owner, introduction state, and prior requests | Private context without permission |
| Open questions | Missing evidence, contradictions, or items requiring confirmation | Confident filler |
| Conversation points | Specific questions that could resolve the open items | A generic interview script |
The format deliberately separates the evidence from the next judgment. An investment committee, principal, or deal lead may weigh the information differently. The teammate's job is to make that decision better prepared, not to make it disappear.
Illustrative example: preparing for a manager meeting
The following example is synthetic.
Northline Family Office is meeting Meridian Ventures about its new early-stage fund. The research teammate receives a narrow assignment:
Prepare a two-page manager brief for Thursday's meeting. Focus on strategy continuity, team changes, portfolio support, and the three unresolved questions from our last conversation. Use current public sources and the authorized Finta relationship record. Cite every material claim. Do not score the manager or recommend an allocation.
The teammate returns:
- a current description of the stated strategy with source dates;
- one team departure found in an official firm announcement;
- a note that the public portfolio page does not show realization data;
- the last meeting's unanswered question about reserves;
- a conflict between the new deck and an older public description of geographic scope;
- four questions for the meeting, ranked by the office's stated review objective.
It also reports: "The strategy-continuity section is ready. I could not verify the claimed realized outcome from an independent or primary source, so I left it as an open question."
That limitation is part of the deliverable. The teammate has reduced reconstruction work without pretending to have completed diligence.
Give the teammate a quality standard
"Research this manager" is too vague for dependable recurring work. Define what a passing brief requires:
- identify the exact entity and current role;
- prefer current primary sources for material claims;
- include an observation date for changing information;
- distinguish public research from authorized relationship history;
- surface conflicts instead of silently choosing one version;
- label inference and missing evidence;
- keep the output within an agreed length;
- end with questions that advance the review objective;
- avoid an investment rating or recommendation unless a qualified human process explicitly owns it.
The NIST AI Risk Management Framework calls for defined tasks, documented roles, and human oversight appropriate to the context. Its AI RMF Core is a useful governance reference for teams assigning continuing work to AI systems. It does not certify any particular research agent.
What still needs a person
An AI research teammate cannot know whether an undisclosed relationship is appropriate to use, whether an internal note remains accurate, or how an investor should weigh a material risk. A person still needs to:
- set the mandate and decision context;
- decide which private sources may be used;
- verify consequential claims and identity matches;
- interpret legal, tax, operational, and investment implications;
- resolve conflicts that the sources cannot settle;
- decide whether to meet, advance, pause, or pass.
Stop the workflow when the subject is ambiguous, a necessary source is inaccessible, confidential information would cross a boundary, or the available evidence cannot support the requested conclusion.
How Aurora can support the responsibility
Aurora can work across available Finta context and supported connections to research, synthesize, and prepare a reviewable next step. A research teammate can use that environment to keep a mandate, working instructions, briefs, corrections, and follow-up questions together over time.
For family offices, the practical goal is not more generated text. It is a repeatable handoff from evidence to a prepared conversation while the principal or investment team retains judgment. See Finta for family offices for the broader relationship workflow, or continue with the AI prospecting teammate when the job is maintaining a qualified target set rather than preparing one subject brief.
Research checked September 27, 2026. This article provides operational education, not investment, legal, tax, compliance, or diligence advice. Verify current sources and apply your organization's review process before relying on a brief.