Alternative assets valuation and portfolio intelligence leader 73 Strings has unveiled Extraction Agent, an agentic artificial intelligence solution engineered to automate financial data collection, extraction, and mapping for private market investment teams.
Built directly into the company’s core data intake engine, 73 Extract, this new feature uses a multi-agent orchestration model which learns from previous changes made by users. Learning from previous changes in a dynamic fashion, the system will reduce repetitive efforts, address formatting inconsistencies, and streamline quarterly portfolio review processes.
Addressing Discrepancies Between Documents and Audits in Private Capital
Portfolio management within private capital markets is an extensive process. Private equity, private credit, and venture capital companies have diversified portfolios that are not bound by any common reporting practices. Dealing with the issue, the deal and operations teams have to spend many hours analyzing different document formats ranging from Excel sheets with multiple tabs to PDFs and other documents as well as earnings comments in order to derive such operational KPIs as revenue, EBITDA, cash balance, debt covenants, etc.
Also Read: Marsh Finalizes Strategic Acquisition of AltamarCAM, Expanding Private Markets Footprint to $75 Billion in Assets Under Management
Such inefficiencies become apparent in each and every quarter and result in the same mapping inconsistencies.
Extraction Agent resolves these systemic issues through a coordinated multi-agent architecture supervised by an intelligent orchestrator. Specialized autonomous agents including dedicated Concierge, Intake, Extraction, and Mapping agents alongside integrated validation services oversee the operational lifecycle from initial ingestion to verified ledger entry.
The platform combines a deterministic data extraction system with a flexible agentic artificial intelligence system. Such a dual-layer structure ensures both the exactness of rule-based systems and flexibility of modern AI solutions in converting data into structured tables without the need for brittle templates.
Main advantages of the solution include:
End-to-End Multi-Agent Solution: Utilizes different agents to process document identification, optical character recognition, table parsing, semantic mapping, and discrepancy verification under a single platform framework.
Continuous Improvement by Soft Memory: Stores all the user feedback, preferred firm taxonomies, and changes in the portfolio mapping. All of these data are stored in the collective memory and used across multiple cycles and multiple users to avoid any potential data mismatch in the future.
Universal Document Processing: Processes documents uploaded via direct platform access or forwarded emails. Documents are linked to their respective portfolio companies and structured metric data is extracted from different sources such as spreadsheets, PDFs, charts, and written information.
Human-in-the-Loop Governance (“AI-as-Maker, Human-as-Checker”): Employs an automated validation engine to identify mathematical discrepancies, low-confidence fields, and outlier trends. Investment analysts review suggested metrics, execute prompt-based bulk updates using natural language commands, and mandate formal human sign-off before data commit.
Direct Source-Level Auditability: Maintains interactive bidirectional lineage, linking every parsed datapoint back to its specific cell, line, or paragraph in the original source documentation to satisfy strict LP and regulatory due diligence audits.
Institutional Security & Zero Data Retention: Ensures client and underlying asset data remains protected under strict enterprise standards, enforcing a strict zero-retention policy that prevents proprietary information from being used to train third-party foundation models.
Executive Commentary and Industry Perspective
“In private markets, the value of any valuation or performance metric depends entirely on trust in the underlying data,” said Yann Magnan, CEO and Co-Founder of 73 Strings. “With LPs and regulators demanding greater transparency, firms need both rapid reporting and absolute auditability. Extraction Agent delivers both by connecting every extracted figure directly back to its source, allowing teams to report faster while standing behind every single number with complete confidence.”
