DEV Community

Cover image for Mapping AI at a venture fund by the data each job touches
PJ (Prakshal Jain)
PJ (Prakshal Jain)

Posted on Originally published at privatesuperintelligence.si

Mapping AI at a venture fund by the data each job touches

Disclosure: I co-founded Mitosis Labs, which runs Private SuperIntelligence. This is an adaptation of the original guide for people who build or run internal tools at venture firms. General information, not legal advice.

If a venture firm asks you to "roll out AI," the hard part isn't picking a model. It's knowing which jobs touch data the firm doesn't own.

What firms actually use it for

Visible's AI Sentiment Report Vol. 1 (2026) surveyed operators, investors and fund managers. It doesn't publish a sample size or survey dates on the report page, so read it as a snapshot:

  • Internal operations: 57% of firms
  • LP reporting: 16%
  • Trends across portfolio data: 16%
  • Benchmarking portfolio performance: 8%

Blockers: security concerns 51%, tool overload 49%, data quality 41%, unclear ROI 16%. Meanwhile 30% of firms adopted zero new AI tools in the past 90 days.

The function-to-data map

ILPA's DDQ 2.0 asks GPs to list the technology tool used per business function. Use the same list as your data classification:

Function Typical AI job Data class
Sourcing Search databases, summarize inbound decks Public, plus synced inbox/calendar
Diligence Read data rooms, draft founder questions Founder confidential, often NDA
IC memos Draft from deck, notes, data room Founder confidential plus partner views
Portfolio monitoring Extract KPIs from updates and board decks Portfolio non-public results
LP reporting / IR Draft letters, DDQ answers LP identities, commitments, side letters
Fund ops Summarize LPAs, track capital calls Fund documents, LP banking details

Only sourcing starts from public data. Everything below it is mostly someone else's confidential information.

Constraints you inherit

  • NDAs. DarrowEverett (June 2026) argues NDAs typically restrict disclosure to "third parties" or a defined "purpose," and putting deal information into an AI tool may breach an NDA "even if an NDA is silent on AI." Treat any external model call with founder data as a disclosure decision.
  • Side letters and the LPA. They set confidentiality for LP names, commitments and terms. Check them before LP data hits any provider.
  • The DDQ. Section 18 asks about security reviews and independent assessments of data held by third parties, and encryption at rest and in transit. Appendix D is a tools template.
  • Plan defaults. OpenAI says personal Free, Plus and Pro accounts may use chats for training unless the user opts out, while Business, Enterprise and Edu workspaces don't train on content by default.

Five decisions to make first

  1. Encode the table above as policy: which data class may reach which tool.
  2. Kill personal-account use for deal work; provision a reviewed workspace.
  3. Get counsel's read on NDA and side letter terms for third-party processing.
  4. Draft the Appendix D tools list and the controls on each.
  5. Require citations on any figure an AI writes into a memo or LP letter.

Tool overload (49%) is the trap here: every single-purpose tool you add for a blocker is another vendor review and another slice of data.

Where we fit

Private SuperIntelligence is run by Mitosis Labs. Each deployment is built for one firm, works across its own email, documents and data, and answers with citations; 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 HVC1-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/ ; ILPA, Due Diligence Questionnaire 2.0 (November 2021), https://ilpa.org/wp-content/uploads/2021/11/ILPA-DDQ-2.0.pdf ; DarrowEverett LLP, Loose Lips, Large Language Models (June 2, 2026), https://darroweverett.com/ndas-non-disclosure-agreements-age-ai-practical-steps-analysis/ ; OpenAI, Data controls in ChatGPT, https://help.openai.com/en/articles/7730893-data-controls-faq ; OpenAI, Enterprise privacy at OpenAI, https://openai.com/enterprise-privacy/

Top comments (0)