Fundraise

Felicis leads $7.5M seed for Unsiloed AI's document tool

What's the deal? Unsiloed AIDealroom has a profile for this one. Try Dealroom → has raised $7.5 million in seed funding led by Felicis to help enterprises make complex documents usable by AI agents. Y Combinator, Orange CollectiveDealroom has a profile for this one. Try Dealroom →, and UncorrelatedDealroom has a profile for this one. Try Dealroom → Ventures joined, along with angel investors Aarthi RamamurthyDealroom has a profile for this one. Try Dealroom → and Beth TurnerDealroom has a profile for this one. Try Dealroom →.

What's the endgame? The San Francisco startup's software breaks down documents with tables, charts, and complicated layouts, then turns them into structured data. It also traces each extracted value back to its source, so businesses can verify output instead of trusting it blindly.

Why now? Enterprises are leaning on AI agents to automate work, but getting information wrong can be costly — especially for a bank, hospital, or law firm. "The problem with LLMs is they're inherently probabilistic," said co-founder and chief executive officer Aman Mishra, pointing to hallucinations that have "a massive negative impact downstream."

Mishra founded Unsiloed with Adnan Abbas, his former classmate at IIT Kharagpur. The pair joined Y Combinator's Fall 2025 batch. Customers already span finance, legal, and healthcare, including voice AI startup Bolna AI.

By the numbers: At $7.5 million, the round sits in the 94th percentile of all-time US fintech seed rounds, drawn from a sample of 8,622 deals.

What could go wrong? The market is crowded. Unsiloed competes with specialists such as ReductoDealroom has a profile for this one. Try Dealroom → and Unstructured, plus offerings from Microsoft and Amazon. It claims 99.78% accuracy on one customer's financial documents, but those results have not been independently verified.

The signal: As companies race to deploy AI agents, the bottleneck is reliable data, not just smarter models. Investors are now backing the narrow but critical task of turning messy enterprise documents into something machines can act on — and prove.

Read more: tezzbuzz.com, TechCrunch

Image credit: Generated with Gemini

More top stories