Automated data extraction uses AI (LLMs) to read and understand the complex, non-standardised schedule of investments from any fund report. This is the only way to eliminate manual data entry and unlock a true, real-time, look-through view of your portfolio’s underlying company exposures, valuations and concentrations.
What is the schedule of investments?
The schedule of investments is arguably the most critical table in a fund’s financial report. It is the detailed, line-by-line breakdown of every investment the fund holds, including:
- Company Name
- Industry / Sector
- Geography
- Quantity / % Ownership
- Cost
- Fair Value / Market Value
- % of Net Assets
It is the raw data for all true portfolio analysis.
Why is this so difficult to process manually?
This is the single biggest source of manual data entry pain for LPs and fund administrators. The problem is a total lack of standardisation.
- Format varies: Every GP uses a different format, with different column orders, names, and structures.
- Column names differ: One fund uses “Fair Value,” another “Market Value,” and a third “Valuation.”
- Data is nested: Assets are often grouped under different strategies or holding companies.
- Footnotes are critical: Key information about valuation methodology or illiquid assets is often in the footnotes.
Old template-based OCR systems fail because it’s impossible to build a template for thousands of different formats, requiring too much input from the users.
How does modern AI decode the schedule?
Modern AI, powered by Large Language Models (LLMs), reads it all just like a human analyst. It doesn’t rely on fixed templates; it understands semantic meaning and context.
- It knows that “Company,” “Investment,” and “Portfolio Holding” all refer to the same thing.
- It can identify the “Fair Value” column, no matter where it is on the page.
- It correctly parses complex, multi-level tables and links footnotes to their corresponding line items.
- It extracts and digitises every single line item from the schedule into a clean, structured database.
What insights does an automated schedule unlock?
Once the data is extracted and aggregated, you move from a stack of static PDFs to a live, dynamic portfolio database. For the first time, you can ask critical questions like:
- True look-through exposure: “What is my total exposure to ‘Acme Corp’ across all 50 funds in my portfolio?”
- Real-time concentration risk: “Show me my aggregate exposure to the UK fintech sector.”
- Valuation tracking: “How has the valuation of ‘XYZ Startup’ changed across all my VC funds over the last 8 quarters?”
- In-depth analysis: “Which of my GPs have the most portfolio overlap?”
This is the how of private market analysis that an intelligent co-pilot is designed to solve.