73 Strings launches Extraction Agent for private-markets reporting

By James Calloway

What did 73 Strings launch?

73 Strings launched Extraction Agent on November 14, 2026; the company says clients of its platform manage more than $20 trillion in assets. Extraction Agent, part of the 73 Extract product, automates financial-data extraction and mapping for portfolio reporting at alternative asset managers.

73 Strings describes document collection as a recurring task for private-markets firms. Those firms receive portfolio-company reports, spreadsheets and financial statements in different formats, according to the company.

The system targets the manual work of collecting and reconciling revenue, EBITDA (earnings before interest, taxes, depreciation and amortization), cash and debt from quarterly company materials used in internal monitoring and investor reporting. It also records user corrections and applies those adjustments to later reporting cycles, the company says.

73 Strings says limited partners (LPs), the investors who provide capital to private-market funds, and regulators demand faster reporting and greater transparency. The launch does not identify any regulator, reporting deadline or audit standard tied to those demands.

Yann Magnan, Chief Executive Officer and Co-Founder of 73 Strings, said in the launch statement:

In private markets, the value of any valuation or performance metric depends entirely on trust in the underlying data.

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.

Magnan’s statement makes trust, rather than raw automation, the product’s central proposition. However, 73 Strings supplies no independent accuracy study or external assurance report with the announcement to support the quoted claim of “absolute auditability.” In plain terms: the company wants to shorten reporting work while keeping a human responsible for every saved number.

How does Extraction Agent process documents?

According to 73 Strings, several software agents—programs assigned to separate steps—divide document handling into six stages:

• Document intake
• Recognition
• Extraction
• Mapping
• Validation
• Review

The company says each extracted figure remains linked to its source document. No dataset can be saved before a user approves it. 73 Strings says those controls create an audit trail (a record of where data came from and what changed) and are intended to reduce repeated manual work.

Extraction Agent can process individual files, batches of reports and emailed documents, 73 Strings says. It identifies the relevant portfolio company and extracts information from tables, charts and commentary across formats and languages, without relying on fixed templates. The company does not specify which languages it supports or explain how conflicting source documents are handled.

Validation checks flag anomalies, inconsistencies and low-confidence values. Users can amend an individual figure or request broader changes through natural-language instructions before approving the final dataset. Those controls could reduce unchecked spreadsheet work, but 73 Strings reports no measured error rate, time saving or reviewer workload in the announcement.

In plain terms: the system makes every number traceable, but traceability is not the same as demonstrated accuracy.

Why is private-markets reporting a target?

The launch reflects a broader effort to apply AI to private-markets middle-office work: data collection, portfolio monitoring and valuation support behind investment reports. Managers in private equity, private credit and venture capital often receive information from portfolio companies in PDFs, scanned files, spreadsheets and other formats, 73 Strings says.

For firms overseeing large portfolios, differing accounting presentations and uneven data quality can compound those problems over repeated reporting periods. In credit strategies, the same issue arises when lenders monitor borrower reports on a recurring basis. These operational frictions make consistent reporting difficult when managers must normalize (convert differently presented data into a consistent structure) without losing its origin.

73 Strings presents inconsistent inputs and recurring review as the core operational problem. It does not quantify document volumes, manual hours or reporting delays, limiting any independent estimate of return on investment.

In plain terms: the product addresses a recurring data chore, though the announcement does not quantify how large that chore is.

What is 73 Strings’ track record?

73 Strings was founded in France in 2020 and now has its headquarters in New York. The company develops valuation and portfolio-intelligence software and says its clients span private equity, private credit, growth equity, venture capital and infrastructure.

Those clients manage more than $20 trillion across public and private markets, according to the company. That figure describes assets managed by clients, not assets managed by 73 Strings itself.

Earlier in 2026, 73 Strings closed a USD 55 million Series B funding round. The company identified the participants this way:

• Goldman Sachs led the round.
• Blackstone Innovations Investments continued its backing.
• Golub Capital, Hamilton Lane, Broadhaven Capital and 7RIDGE participated.

The company operates offices in New York, San Francisco, London, Paris, Riyadh, Abu Dhabi, Singapore, Bengaluru and Hyderabad. Its software focuses on data extraction, standardisation, portfolio monitoring and valuation workflows for institutional investors and managers. The funding and geographic footprint show institutional reach, not measured Extraction Agent performance.

In plain terms: 73 Strings has capital, customers and a broad footprint, but those facts cannot establish how well the new agent works.

What remains unproven?

73 Strings’ launch report details workflow features but remains silent on several commercial and technical measures. It does not disclose:

• Accuracy or error-rate benchmarks
• Measured reductions in reporting time
• Implementation schedules or staffing requirements
• Pricing and contract terms
• Named customers or quantified case-study results
• Independent validation of its speed and auditability claims

Source-document links and required human approval are meaningful safeguards, yet they do not prove that extracted values are correct. A reviewer can approve a flawed figure if the extraction or source material was wrong, although the announcement does not quantify that risk. The same silence applies to how much review the automation actually saves.

The stakes are practical. Managers need reports quickly enough for LPs and regulators, but speed will matter only if faster outputs remain defensible when challenged. 73 Strings has not yet supplied the comparative error and time data needed to judge whether its automation improves on the manual process it aims to replace.

In plain terms: the next proof point is measured accuracy, customer results and time savings that preserve the required human sign-off.