New platform architecture enables firms to direct Lumonic using natural-language instructions, beginning with automated data extraction from portfolio company and borrower documents

Lumonic, a PitchBook company, today launched Lumonic 12.0, a platform update that enables firms to direct Lumonic using natural-language instructions. The platform update ensures governance and traceability by recording every action an agent takes, requiring human approval before any data is written, and tracing every value back to its source document. The release begins with data extraction, which reads documents from portfolio companies and borrowers and converts them into structured, source-traceable data automatically, reducing the manual re-entry that has slowed portfolio monitoring teams. AI extraction is the first of several AI workflow capabilities Lumonic plans to release over the next few months on its new agent platform, designed to support every stage of the portfolio monitoring workflow.

Why It Matters

Portfolio monitoring in private markets runs on unstructured documents. Portfolio companies report inconsistently, formats and reporting periods vary, and line items change quarter to quarter, requiring monitoring teams at institutional credit and private equity firms to consistently re-key figures into models manually, a process that is slow and difficult to audit. What’s more, firm-specific context is often retained by individual analysts rather than the system itself, requiring manual review to reconcile inconsistencies. Structured data still typically requires a dedicated reporting interface to query, adding time to routine requests. With portfolio reviews commonly conducted on a quarterly basis, issues such as covenant breaches or missed budgets may go unaddressed for weeks.

As adoption of AI tools across the financial industry increases, the reliability of underlying data increasingly determines the reliability of AI-powered output. Lumonic 12.0 is designed to pair the speed and efficiency of AI with the accuracy and traceability portfolio monitoring requires.

Lumonic 12.0

Lumonic's new platform architecture is designed to support automation across the broader monitoring workflow, enabling users to query the platform with natural language search. Every action taken on the platform is recorded in a durable, auditable log, allowing teams to review each step of a process rather than only the outcome. Decisions a firm makes are retained and applied automatically going forward, so the same reporting pack no longer produces different numbers depending on which analyst reviews it. New automated workflows are currently configured in partnership with the Lumonic team today, with self-directed configuration planned for a future release. Looking ahead, Lumonic will also introduce event-triggered automations that flag late reporting, covenant breaches, and portfolio changes as they occur, along with report builders, to close the gap between when a document lands and when a firm can act on it.

"Portfolio monitoring has always come down to trust: can a firm stand behind a number when a limited partner asks where it came from?" said Kevin Hsu, Co-founder and CEO of Lumonic. "Lumonic 12.0 makes the source of every figure visible, the same way an analyst would show their work. Our customers already rely on Lumonic for an AI-first approach to portfolio monitoring, and with 12.0 we're extending that foundation into automation that takes on the work that has kept teams from mission-critical decisions. Alongside PitchBook, a firm can hold its own numbers and the broader market to one standard, in one place."

How AI Extraction Works

As the first fully automated workflow, AI extraction automatically collects, analyzes, and reports portfolio performance, with each figure traceable to its exact location in the source document, allowing teams to audit results without repeating the underlying work. Decisions a firm makes once, such as how it defines a metric or treats a new line item, are applied automatically in subsequent periods. More information on how the capability works is as follows:

• Automatic start: A document is uploaded or forwarded to Lumonic, and the extraction agent begins working without manual setup.
• Two modes: For a new company, the agent builds the extraction structure from the underlying documents. For a returning company, it recognizes the reporting period, carries forward prior decisions, and flags only what has changed.
• Built-in checks: Extraction rules review the work before a person sees it, including balance sheet tie-outs, cash flow reconciliation, and materiality thresholds for restatements. When a rule fails, the agent attempts to self-correct and flags anything it cannot resolve.
• Full traceability: Every extracted value links back to the exact cell or page in the source document, with the same traceback available in Lumonic's Excel plugin and over MCP.
• Human-approved publishing: New extractions default to ready for review, with auto-publish available as an opt-in setting. Edits can be made by hand or described in plain language, and corrections carry forward to future extractions automatically.

PitchBook and Lumonic

Lumonic was acquired by PitchBook, a Morningstar company, in 2025. The acquisition connects portfolio monitoring and private market intelligence under one company, giving investment professionals a way to manage and analyze their own portfolios alongside PitchBook's private capital data. This release follows Lumonic's direct MCP release earlier this year, further demonstrating Lumonic's AI-first innovation roadmap and addressing critical workflow inefficiencies in portfolio monitoring. Since the acquisition, Lumonic has more than doubled its customer base year-over-year and supports management of more than 5k portfolio companies.

"Two generations of portfolio monitoring software made the tooling better while the workflow stayed human. Someone still opens the document, interprets it, and decides what matters," said Andrew Riess, Co-founder and Head of Product and Engineering at Lumonic. "12.0 is the first release where the platform does that work itself. It runs from your firm's own definitions, applies the decisions your team already made, and traces every number back to the page. The foundation took years to build."

For more information about Lumonic 12.0, click here.

About PitchBook, a Morningstar company

As the pulse of private capital markets, PitchBook delivers trusted, real-time data, research, and technology to help investors, dealmakers, and innovators make decisions with confidence. Its products provide comprehensive information on companies, investors, funds, deals, and people, along with tools that help professionals analyze market activity and make informed decisions. Founded in 2007, PitchBook today serves more than 100,000 clients worldwide and is recognized as the leading source of private capital market intelligence. PitchBook has grown to over 3,000 employees across offices in Seattle, San Francisco, New York, London, Singapore, Mumbai, and other global locations. Since 2016, PitchBook has operated as a subsidiary of Morningstar, Inc.

View source version on businesswire.com: https://www.businesswire.com/news/home/20260917410818/en/