Disclosure: I co-founded Mitosis Labs, which runs Private SuperIntelligence. This is an adaptation of the original guide written for venture investor relations teams. General information, not legal advice.
If you build internal tools at a venture fund, the quarterly LP letter will land on your desk as "can the AI write this?" It can draft it. The engineering problem is making every number in that draft checkable, and keeping the model away from the numbers it should never produce.
Adoption is uneven. Visible's 2026 AI Sentiment Report found internal operations lead at 57% of firms, while LP reporting sits at 16%. Security concerns (51%) and data quality (41%) are among the blockers it lists. Both are design problems you can address.
Inputs and the target schema
Inputs are messy: founder update emails, board deck PDFs, spreadsheets and the fund administrator's statements. Collection tooling has moved fast here. Visible reports its AI Inbox share rose from 10% to over 40% of metrics extracted, and Standard Metrics says its metrics are "traceable to source documents."
The target is less messy than it looks. ILPA's Reporting Template (v. 2.0, released January 22, 2025) and its Performance Template give you a field-level structure, including standardized reporting of IRRs and TVPI/MOIC. Treat them as your output schema and your review checklist.
The pipeline
-
Extract. For each portfolio company, produce rows of
metric, value, period, document, page, definition_used. - Normalize. Map to the fund's metric definitions. Emit a flag when a company's definition differs from last quarter.
- Inject fund figures, never compute them. TVPI, DPI and IRR come from the fund admin or accounting system as quoted inputs with a source reference. The model does no fund-level arithmetic: one error there reaches every LP at once.
- Draft. Generate commentary per company from the fixed table only. Every number carries a citation; anything inferred rather than read gets a marker.
- Review gates. CFO signs off numbers, the responsible partner signs off narrative.
- Send gate. No LP communication leaves without a human approving it. Make it a permission, not a convention.
A useful invariant for tests: every numeric token in the draft must resolve to a row in the extraction table or the fund-admin input. Fail the build otherwise.
Data boundaries
LP reporting touches two confidential sets: portfolio companies' non-public results and the LPs' own information. Practical rules:
- Keep LP identities and capital accounts out of the drafting context unless a task needs them. Most commentary needs portfolio data only.
- Check the plan's training and retention terms with your provider directly, and know who at the provider can read content.
- Check the LPA and side letters. Confidentiality terms may treat sending LP information to a vendor as disclosure to a third party.
- Be ready to describe it. ILPA's DDQ 2.0 asks about reporting and investor relations performed by third parties, and about security reviews of firm data held by third-party providers. ILPA's Responsible AI Quick Guide (January 2024) suggests LPs ask, "What are the GP's controls around AI use at the fund level?"
Regulatory note
The SEC's quarterly statement rule no longer applies: on June 5, 2024, the Fifth Circuit vacated the private fund adviser rules in full. Quarterly reporting terms now come mainly from the LPA, side letters and market practice such as ILPA's templates.
Where we fit
Private SuperIntelligence from Mitosis Labs works across a firm's own documents and inboxes, answers with citations a CFO can check, and acts within permissions the firm sets, so a person approves anything before it is sent. Each deployment is built for one firm, and the provider keeps no copy of the work and trains nothing on it.
Speak with the founders
Private SuperIntelligence takes a limited number of families at a time so every one gets the same standard of service. Book a private call (ref LPR1-DEV) to talk through your office with Alex Morris.
Prakshal Jain (PJ), Co-founder and CTO, Private SuperIntelligence (by Mitosis Labs)
Alexander Morris, Co-founder and CEO, Private SuperIntelligence (by Mitosis Labs)
Sources: Visible, AI Sentiment Report Vol. 1 (2026), https://visible.vc/blog/ai-sentiment-report-v1/ ; Standard Metrics, https://standardmetrics.io/ ; ILPA, Updated Reporting Template and New Performance Template (January 22, 2025), https://ilpa.org/news/ilpa-releases-updated-reporting-template-and-new-performance-template-for-industry-adoption/ ; Morgan Lewis, Fifth Circuit Vacates SEC Private Fund Adviser Rules in Full (June 2024), https://www.morganlewis.com/pubs/2024/06/fifth-circuit-vacates-sec-private-fund-adviser-rules-in-full ; ILPA, Due Diligence Questionnaire 2.0, https://ilpa.org/wp-content/uploads/2021/11/ILPA-DDQ-2.0.pdf ; ILPA, Responsible AI Quick Guide for Asset Owners (January 2024), https://ilpa.org/wp-content/uploads/2023/10/Responsible-AI-Quick-Guide-for-Asset-Owners.pdf
Top comments (0)