Should Limited Partners use AI to assist with investment decisions?

Private markets have grown from a niche to a central pillar of institutional portfolios, yet their decision-making infrastructure remains analog. Limited partners (LPs) must allocate billions based on incomplete, delayed, and strategically curated disclosures, while managing decade-long feedback loops. At the same time, artificial intelligence (AI) and machine learning (ML) are revolutionizing how information is parsed, interpreted, and acted upon. This paper reviews early evidence that algorithms can read GP reports, forecast cash flows, and benchmark funds. But it also shows where the limits lie.

Limited Partners vs Unlimited Technologies:
How Tech Could Transform Investing in Private Capital Funds

• Braun, Tamayo, López-de-Silanes, Phalippou, and Sigrist
• Working paper, 2025
• A version of this paper can be found here
• Want to read our summaries of academic finance papers? Check out our Academic Research Insight category

Key Academic Insights

Private Equity’s Data Problem Defines the AI Challenge
LPs face delayed, fragmented, and strategically framed disclosures often – narrative-heavy and unstandardized. AI models trained on high-frequency public data cannot simply be transplanted into private markets without accounting for these structural differences.

Textual Analysis Unlocks Predictive Signals in Disclosures
Recent studies show that fundraising documents and GP interim reports contain forward-looking information. Natural language processing (NLP) and supervised learning can extract predictive signals -tone, sentiment, and narrative structure – that correlate with future performance, even after controlling for reported valuations.

Large Language Models Offer Speed, Not Omniscience
LLMs can summarize private placement memoranda, extract key clauses, and standardize ESG reports, but they cannot yet link language to future outcomes. They accelerate workflows and interpretation but require human validation and economic reasoning to avoid bias, hallucination, or overconfidence.

Governance and Confidentiality Remain Binding Constraints
Some general partners now restrict LPs from processing documents through AI tools due to confidentiality and data security concerns. This emerging conflict underscores that technological adoption in private equity depends as much on institutional and legal frameworks as on technical capability.

Practical Applications for Investment Advisors

Treat AI as an interpretive tool, not an oracle.
Technology can help LPs process and compare disclosures efficiently, but decisions still require domain expertise. The edge lies in judgment – in knowing when to trust, challenge, or contextualize model outputs.

Build governance for model oversight.
AI adoption demands protocols for validation, explainability, and accountability. LPs must ensure that decisions influenced by models remain transparent and auditable, especially under fiduciary scrutiny.

Integrate narrative analysis into due diligence.
The language of GP disclosures -how a manager articulates strategy, risk, and governance – contains measurable predictive content. Advisors should incorporate textual analysis into fund evaluation frameworks to capture these qualitative signals.

Prepare for contractual friction.
As confidentiality restrictions tighten, advisors may need to negotiate explicit rights to apply AI to fund documents or develop secure, in-house analytics frameworks. The evolution of “AI clauses” in partnership agreements may become a new frontier in fund governance.

How to Explain This to Clients

“Private equity investing has always been about trust and access—but that world is changing. Machines can now read fund documents and detect patterns that humans overlook. Still, technology is no silver bullet. The best investors will combine machine-driven efficiency with human oversight, ensuring that AI insights are interpreted with experience, not replaced by it.”

The Most Important Chart from the Paper

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged and do not reflect management or trading fees, and one cannot invest directly in an index.

Abstract

We examine how artificial intelligence and machine learning may alter decision-making in private markets. Unlike public equity, where frequent and standardized disclosures enable rapid validation of predictive models, private markets are defined by sparse, delayed, and strategically framed disclosures. This mismatch is the central tension between limited partners and unlimited technologies: investors face opacity and decade-long feedback loops, while algorithms are designed for abundant and high-frequency signals. Recent evidence shows that textual features of fundraising documents and GP reports contain systematic predictive content, that deep learning methods can improve cash flow forecasting, and that benchmarking approaches can realign fund categories with underlying exposures. At the same time, data scarcity, model interpretability, and confidentiality restrictions remain binding. We outline a research agenda that integrates computational methods with economic reasoning, emphasizing oversight, causal inference, and transparency as preconditions for reliable adoption.