Key Takeaways
- • 73 Strings, an AI-powered valuation and portfolio intelligence platform for alternative asset managers, launched Extraction Agent, an agentic AI solution that automates financial data extraction and mapping from portfolio company reports.
- • Extraction Agent uses a multi-agent workflow (concierge, intake, extraction, mapping and validation agents) that combines a deterministic extraction core with an agentic AI layer, requiring no rigid templates.
- • The system's Soft Memory learns from user corrections and applies them to future documents, which can be shared across teams to reduce repetitive manual work.
- • Every extracted figure links back to its location in the source document, with human sign-off required before data is saved, addressing LP and regulator demands for auditability.
- • Target use cases include extracting revenue, EBITDA, cash and debt from quarterly reports and automating borrower reports for credit monitoring in private equity, private credit and venture capital.
Summary
73 Strings announced the launch of Extraction Agent, a new agentic AI solution built into its 73 Extract product, according to a Business Wire release. The tool is designed to address a significant operational burden for alternative asset managers, whose investment teams must reconcile inconsistent financial statements, spreadsheets and quarterly reports across portfolios of companies with varying reporting formats. The company said these formatting inconsistencies and data mapping issues recur across reporting cycles.
Extraction Agent relies on a coordinated multi-agent workflow orchestrated by an intelligent agent orchestrator. The concierge, intake, extraction, mapping and validation capabilities manage the process from document intake and recognition through extraction, mapping, validation and review. The company describes the design as combining the reliability and consistency of a deterministic extraction core with the adaptability of an agentic AI layer, allowing complex financial documents to be converted into structured, review-ready data while maintaining source-level traceability.
A central feature is Soft Memory, which retains user updates, reporting preferences and data mappings and applies them to subsequent documents. The company says Soft Memory improves with use and can be shared across teams. The product accepts individual or multiple reports, including emailed files, and processes spreadsheets, PDFs, scanned documents and images in multiple languages. A dedicated validation agent applies built-in rules and confidence checks to flag anomalies and low-confidence values. Users can correct individual values or request broader changes through natural-language instructions, and human sign-off is required before data is saved. The company describes this as an "AI-as-maker, human-as-checker" approach.
CEO and Co-Founder Yann Magnan said in the release that "in private markets, the value of any valuation or performance metric depends entirely on trust in the underlying data," adding that LPs and regulators are demanding greater transparency. The company states that zero data retention applies and that no client data is used for AI model training.
The product is aimed at private equity, private credit, venture capital and other alternative strategies. 73 Strings will host a live product demonstration on Thursday, December 3 at 10:00 AM ET. The company, founded in France in 2020 and headquartered in New York, closed a $55 million Series B in February 2025 led by Goldman Sachs, with participation from Blackstone Innovations Investments, Golub Capital, Hamilton Lane, Broadhaven Capital and 7RIDGE. The release states the platform supports clients managing more than $20 trillion in assets, a figure the company attributes to its client base rather than to this product.
Extraction Agent relies on a coordinated multi-agent workflow orchestrated by an intelligent agent orchestrator. The concierge, intake, extraction, mapping and validation capabilities manage the process from document intake and recognition through extraction, mapping, validation and review. The company describes the design as combining the reliability and consistency of a deterministic extraction core with the adaptability of an agentic AI layer, allowing complex financial documents to be converted into structured, review-ready data while maintaining source-level traceability.
A central feature is Soft Memory, which retains user updates, reporting preferences and data mappings and applies them to subsequent documents. The company says Soft Memory improves with use and can be shared across teams. The product accepts individual or multiple reports, including emailed files, and processes spreadsheets, PDFs, scanned documents and images in multiple languages. A dedicated validation agent applies built-in rules and confidence checks to flag anomalies and low-confidence values. Users can correct individual values or request broader changes through natural-language instructions, and human sign-off is required before data is saved. The company describes this as an "AI-as-maker, human-as-checker" approach.
CEO and Co-Founder Yann Magnan said in the release that "in private markets, the value of any valuation or performance metric depends entirely on trust in the underlying data," adding that LPs and regulators are demanding greater transparency. The company states that zero data retention applies and that no client data is used for AI model training.
The product is aimed at private equity, private credit, venture capital and other alternative strategies. 73 Strings will host a live product demonstration on Thursday, December 3 at 10:00 AM ET. The company, founded in France in 2020 and headquartered in New York, closed a $55 million Series B in February 2025 led by Goldman Sachs, with participation from Blackstone Innovations Investments, Golub Capital, Hamilton Lane, Broadhaven Capital and 7RIDGE. The release states the platform supports clients managing more than $20 trillion in assets, a figure the company attributes to its client base rather than to this product.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.