Funding & VC

Unsiloed AI raises $7.5M to make documents usable by AI agents

Unsiloed AI raised $7.5 million in seed funding led by Felicis. The seven-person startup serves finance, legal and healthcare clients and prices document extraction from one cent per page.

By Grace Kim

4 min read

Felicis leads $7.5M bet on Unsiloed AI to make enterprise documents usable by AI agents
Felicis leads $7.5M bet on Unsiloed AI to make enterprise documents usable by AI agentsClickerHappy / pixabay

What's News

  • Unsiloed AI raised $7.5 million in seed funding led by Felicis.
  • A test on 542 financial documents reported 99.78% accuracy across roughly 84,000 extracted data points; the results were not independently verified.
  • In startup-run olmOCR-Bench tests, Unsiloed scored 88.0 versus 66.0 for Reducto and 39.9 for Unstructured across 1,403 PDF pages.
  • Unsiloed’s pricing starts at one cent per page.
  • The startup joined Y Combinator’s Fall 2025 batch and currently has seven employees.

By Grace Kim

Unsiloed AI has raised $7.5 million in seed funding led by Felicis to build software that turns complex enterprise documents into structured data for AI agents.

According to the startup, the investors were:

  • Felicis, a Silicon Valley venture capital firm and the lead investor
  • Y Combinator
  • Orange Collective
  • Uncorrelated Ventures
  • Angel investors Aarthi Ramamurthy and Beth Turner

Why do complex documents still trip AI agents?

Unsiloed AI says businesses are turning more work to AI agents, but tables, charts and complicated layouts can still confuse those systems. Mishra warned that errors can carry severe consequences for banks, hospitals and law firms, which rely on complex documents in their daily work.

Unsiloed co-founder and CEO Aman Mishra said:

“The problem with LLMs is they’re inherently probabilistic. There is going to be some deviation induced. There is going to be some hallucination involved, and that has a massive negative impact downstream.”

The company’s software breaks down complex documents, extracts information from tables, charts and other elements, and produces structured data that agents can use. It also shows where each extraction came from in the original document, so businesses can check results instead of taking AI output at face value.

Businesses can check what the AI extracted instead of trusting a plausible answer.

How does Unsiloed process complex documents?

Mishra said the startup uses “dual-stream” vision models that handle text and image data separately, paired with “heat map-based chunking” to identify a document’s most important parts.

The system then directs more computing power toward figures, numbers and charts. A specialized decoder accounts for industry-specific terminology and helps distinguish terms that carry different meanings in different business contexts.

“In one of the streams you have the text tokens, in the other stream you have the exact image tokens,” Mishra said. “This is something that we have spent a lot of time architecting.”

Mishra said the approach also helps keep Unsiloed’s costs down by directing computing power toward the parts of a document that need the most attention, instead of processing every section in the same way.

In plain terms, Unsiloed spends more compute on the hardest parts of a page.

How strong are the accuracy claims?

Mishra pointed to a test involving 542 financial documents supplied by an enterprise customer. In a technical brief shared by the startup, Unsiloed reported 99.78% accuracy across roughly 84,000 extracted data points.

The test compared Unsiloed with major AI models, Microsoft’s Azure Document Intelligence and the customer’s in-house extraction tool. The results have not been independently verified.

Unsiloed competes with Reducto and Unstructured, document-processing offerings from Microsoft and Amazon, and general-purpose AI models that businesses use to extract information.

On the startup’s tests for olmOCR-Bench, a public benchmark developed by Allen AI, Unsiloed scored 88.0. Reducto’s agentic system scored 66.0, and Unstructured scored 39.9. The benchmark covered 1,403 PDF pages and tested tables, mathematical expressions, scanned documents and complex layouts.

The company’s benchmark scores came from tests it ran itself, and its 99.78% figure has not been independently verified.

Who is using Unsiloed, and what will the seed money fund?

Mishra said Unsiloed already works with customers in finance, legal services and healthcare. One unnamed bank uses the technology to process documents. An unnamed law firm has deployed it on its own servers to handle contracts and old case files.

Bolna AI is the only customer Mishra named. The voice AI startup uses Unsiloed to ingest knowledge bases for building voice agents.

Unsiloed is a San Francisco-headquartered startup co-founded by Mishra and Adnan Abbas, Mishra’s former classmate at the Indian Institute of Technology (IIT) Kharagpur. The two joined Y Combinator’s Fall 2025 batch after developing the technology.

Pricing starts at one cent per page, Mishra said. Unsiloed currently has seven employees and plans to use the seed money to hire more researchers and engineers, expand sales efforts and develop proprietary datasets to improve its technology.

The next test is whether Unsiloed can turn deployments at a bank, a law firm and Bolna AI into broader sales while expanding its seven-person team.

Original: unsiloed.ai

Share this article:

More from Grace Kim

Grace Kim

Show full bio

Market editor covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.

2,301 articles

Related articles

« Previous articleNext article »