PE-AI evidence-linked decision support
Contact
PE-AI · Private Equity Decision Support

Evidence-first AI for private equity.

Turn investment information into traceable decisions.

PE-AI is an AI decision-support platform designed to accelerate due diligence and post-investment risk monitoring while keeping evidence and human judgment in the loop. Every finding is built to stay connected to the source material behind it.

Domain-specific AI Evidence-linked Human-supervised
25,000+
venture and PE institutions in China
150,000+
investment professionals
RMB 15T+
assets under management in the market
The problem

Private equity decisions start with fragmented evidence.

Analysts must turn scattered documents and fast-moving information into high-stakes judgments under time pressure. Generic AI can write fluent answers, but it rarely makes the evidence inspectable.

01

Fragmented information

Investment data lives across pitch decks, financial statements, contracts, reports, email and external sources, with no shared standard to tie it together.

02

Repetitive analysis

Screening, extraction, cross-checking and report drafting consume time that should go to judgment, founder conversations and relationships.

03

Opaque AI outputs

Generic tools produce confident answers without an easy way to inspect the source, challenge a finding, or assign a named reviewer.

Field observation: more than 90% of domestic PE firms still rely on spreadsheets, instant messaging and closed legacy systems for research and management. The figure is drawn from the team's interviews with 25 PE institutions, internal data and industry databases.
The product

Start with due diligence, then expand across the investment lifecycle.

PE-AI transforms fragmented investment information into structured, evidence-linked analysis — a PE-specific intelligence layer rather than a prompt on a general model.

Step 1

Ingest

Documents and data from deal and research sources enter a unified standard.

Step 2

Analyze

Extract key facts and figures from document-heavy material.

Step 3

Identify

Surface risks and signals against historical project data.

Step 4

Trace

Link every finding back to its evidence and location.

Step 5

Review

A named professional verifies before any decision is made.

Initial focus: AI-powered due diligence

From document-heavy review to a structured, reviewable investment work product.

  • BP screening and document intake
  • Fact extraction and cross-checking
  • Evidence-linked report drafting

Adjacent workflow: post-investment monitoring

The same evidence-first approach extends beyond the deal decision.

  • Operating and financial signal monitoring
  • Emerging risk flags and early warning
  • Periodic review summaries

Output: structured diligence notes · risk flags · evidence-linked report drafts

Two flagship applications

Where the evidence-first approach is already being tested.

Both workflows are built on the same principle: use only information available at the time of decision, and output a complete, traceable decision chain.

Scenario 01 · Risk control

Intelligent risk review engine

The engine parses a full investment memo, checks document completeness, data consistency and the reasonableness of key indicators, then matches potential risks against a standardized rule base built from historical projects.

Review cycle
Hours to 10 seconds
Common errors caught
95%
Institutions testing
3 PE firms
Input
Investment or IC report uploaded as the source for review.
Processing
Knowledge-graph scan across completeness, consistency and logic.
Output
A structured issue list that can be exported and acted on directly.
Scenario 02 · Investment committee

IC decision briefing

The system consolidates every document required for an investment committee — memos, due diligence, risk reports and financial models — and produces a concentrated one-to-two-page briefing that structures the core logic, highlights, risks and key figures.

Source material
100+ pages
Processing time
10 seconds
Reading time
3 hours to 10 minutes

Covered across the full lifecycle: BP screening, due-diligence report generation, legal document processing, 7×24 post-investment monitoring with roughly three-month early warning, and an institutional knowledge repository that lets new staff query why a past decision was made.

How it works

A PE-specific intelligence layer, not a prompt on a general model.

The architecture follows the research-validated SSFF multi-agent framework, organized around PE workflows and evidence retrieval rather than a single generic large model.

Application layer

Investment workflows

Deal screening, due diligence, investment reports and post-investment monitoring, delivered across web and mobile with fine-grained permissions.

Agent & reasoning layer

SSFF multi-agent collaboration

Task decomposition, specialized agent roles and external knowledge retrieval simulate the front-office, middle-office and decision functions of a real PE team.

Knowledge layer

PE knowledge graph + GraphRAG

Entities, projects, people and events form a graph database with LLM-assisted retrieval, grounding reasoning in the institution's own history.

Data & security layer

Integration and controlled deployment

Standard APIs, enterprise permissions and hybrid deployment keep raw data in-domain while supporting privacy-preserving model training.

Source-linked findings

Each conclusion stays connected to the evidence behind it, so analysts can inspect and challenge it.

Permission-aware access

Fine-grained controls decide who can see which documents and findings.

Human review before decisions

AI assists with analysis and documentation; professionals retain the investment judgment.

Performance

Measured outcomes in real workflows.

3
PE institutions running internal tests
80%
less repetitive work in tested workflows
98%
information extraction accuracy on real PE data
95%
common risk-control errors identified in review
Recognition and intellectual property

Research origins meeting an industry incubation pathway.

Industry incubation

Qianhai official AI incubation program

Selected from nearly 1,000 global applicant teams for the Shenzhen Qianhai Authority's official AI industry incubation program, with an industry-investment panel presentation completed.

IP portfolio

Two patents

  • Patent · LP-GP privacy matching and due-diligence risk control
  • Patent · Streaming post-investment risk early warning and exit support

Patent applications are published and pending; current legal status is confirmed before external disclosure.

Journey

From a university research project to a tested prototype.

Mar 2024

Project launched

Started as a national-level undergraduate innovation project.

Sep 2024

Research output

Contributed to a national-level textbook on AI and venture capital.

Mar 2025

Core R&D

Algorithm development and theoretical validation completed.

Dec 2025

Prototype V1.0

Prototype completed and institutional testing initiated.

2026

Incubation and IP

Selected for the Qianhai incubation program; patent applications published.

Team

Built at the intersection of AI research and PE workflows.

An interdisciplinary core team

The PE-AI algorithm R&D group brings together artificial intelligence, computer science and engineering, and economics and finance, combining a technical base with quantitative analysis and first-hand exposure to primary-market workflows.

Academic, industry and professional support

A three-part support system combines senior scholars leading national and provincial research funds, experienced PE partners shaping product requirements, and a professional network of lawyers, certified accountants and financial advisors.

South China University of Technology
Contact

Request the full demo, test reports or research collaboration.

The team is happy to share the working prototype, internal testing materials and the technical detail behind the evidence-linked architecture.

contact@usepeai.com