To minimize the operational burden that investors face managing portfolios of private companies, Standard Metrics provides a document data parsing service that surpasses traditional internal processes in accuracy, speed, and centralization.

Historically, we’ve ensured high data accuracy and traceability through our U.S.-based team of skilled analysts, supported by in-house workflow software. This precision is essential, as the data we onboard is used by our customers for audits, LP reporting, valuations, and other mission-critical functions.

As we scale, our goal is to deliver this data faster and at higher volumes — without compromising accuracy — so that firms can access more insights about their portfolio, more quickly. AI holds great promise for streamlining this effort.

Our customers have tested other off-the-shelf AI solutions and found them lacking in accuracy. Alternative approaches, like document mapping, have also fallen short, proving time-intensive and requiring frequent rework as financial statement formats evolve.

Standard Metrics has pioneered a new solution: an AI agent that supports, rather than replaces, a managed data services team to ensure efficiency and accuracy.

What we built

We’ve moved from a human-centered document parsing process to a multi-step parsing flow built to improve speed of parsing while maintaining high accuracy.

To support this flow, we first rely on pre-processing to split large documents into smaller parts and classify those smaller documents by type (balance sheets vs. income statements as well as PDFs vs Excel files, for example). More on why can be found here. We then have the LLMs take a first pass at parsing these simplified documents. Our managed data services team then can QA, edit, or add to the LLMs’ work inside of our application. Along the way, we also measure accuracy of our AI-parsed metrics on a continuous basis and track errors over time through a rigorous evaluations process.

Our AI parsing effort has grown rapidly over the past couple of quarters and now handles a significant, double-digit percentage of all data points parsed by our managed data services team. Each human analyst on our team is able to ingest more data, shifting some of their time from ingesting data to overseeing our AI agents’ work. We expect this trend to continue moving forward. This means we can parse more customer data, more quickly and help customers save on overhead costs from manual and inefficient internal document parsing. We’ve also been able to maintain our aggressive internal standard for data accuracy by scaling AI data parsing with close human supervision and QA.

What’s next?

Beyond continuing to improve AI-powered data parsing, the Standard Metrics team is also increasingly leveraging AI to make portfolio monitoring and reporting easier for our customers. A few things we’re designing and building include:

• Natural language questions & answers: Ask the questions you need from your data in natural language (e.g. “What is the cash balance for this company over the past 5 quarters?” or “What are some companies that might compete with this company?”) rather than complicated queries, and get natural language responses back.
• Custom reporting: Build unique visuals and reports in-application with natural language queries rather than SQL.
• Summarization: Finding the important key topics and trends across quarters of data and notes to make reporting faster.

We’re excited to embrace the speed and reasoning capabilities of AI, while maintaining a component of human supervision where consistently high accuracy is paramount. Get in touch via the form below if you’d like to learn more about how we can help automate your portfolio reporting process.