The Investor Answer Is Only Correct Inside the Right Boundary

An accurate answer shown to the wrong investor is still a serious failure.

AI-assisted investor reporting has two jobs: assemble the answer and preserve the boundary around it. The second job is not a prompt instruction. It belongs in authorization and data design.

Scope before retrieval

Before an agent searches, the system should know:

• the authenticated investor;
• the funds and entities that investor may access;
• the applicable reporting periods;
• the documents released to that investor;
• any side-letter terms relevant to the request.

Retrieval happens inside that scope. The interface should not fetch a broad result and hide unauthorized rows afterward.

A useful answer carries its record

An LP response should preserve the source document, section, reporting date, and release status behind each material statement. If a draft letter and a published statement disagree, the conflict should be visible to investor relations.

Separate drafting from publishing

An agent can prepare:

• an answer to a fee or allocation question;
• a quarterly letter draft;
• a statement package;
• a summary of prior calls and commitments;
• a document request response.

A named investor-relations owner should approve anything external. Draft access and publication access should be separate permissions.

Test the negative case

A permissions test should not only prove that Investor A can see Investor A's records. It must prove that Investor A cannot request Investor B's record by changing an identifier, a URL, or the wording of a question.

The same test belongs at fund and client boundaries.

Make commitments visible

The answer often lives outside a formal statement. Prior calls, letters, and side letters can contain commitments that shape the response. An agent can surface those records, but the firm still decides which commitment governs.

Fast investor service matters. Correct scope matters more.