Felicis leads $7.5M bet on Unsiloed AI to make enterprise documents usable by AI agents
With a team of just seven, Unsiloed AI plans to compete with tech giants and specialized startups in the crowded AI document-processing market.
Businesses are increasingly relying on AI agents to automate tasks, but unlocking the information stored in complex documents remains a significant hurdle. To address this issue, Unsiloed AI has secured $7.5 million in seed funding, led by venture capital firm Felicis. The technology aims to make enterprise documents more accessible to AI agents through specialized software.
Felicis led the round, with participation from Y Combinator, Orange Collective, Uncorrelated Ventures, angel investors Aarthi Ramamurthy and Beth Turner.
Documents often contain intricate layouts, tables, and charts that AI systems struggle to interpret accurately. Misinterpretations can lead to costly errors, particularly for businesses operating in finance, healthcare, and legal services. Unsiloed aims to tackle this problem by breaking down complex documents, extracting information from tables, charts, and other elements, and converting it into structured data that AI agents can utilize.
The technology also traces the source of the extracted information, enabling businesses to verify the results rather than relying solely on AI-generated output.
Co-founders Aman Mishra and Adnan Abbas, former classmates at India's IIT Kharagpur, developed the technology and joined Y Combinator's Fall 2025 batch to further refine it. Unsiloed is already collaborating with customers in finance, legal services, and healthcare. A bank is utilizing the software to process documents, while a law firm has deployed it on its servers to handle contracts and historical case files. Voice AI startup Bolna AI also uses Unsiloed to extract knowledge from its databases for voice agents.
Although Unsiloed operates in a market already populated by specialized document-processing firms like Reducto and Unstructured, as well as offerings from tech giants Microsoft and Amazon, its approach sets it apart. The company employs "dual-stream" vision models that handle text and image data separately, along with "heat map-based chunking" to identify crucial sections of a document.
This strategy directs more computing power toward essential elements such as numbers, figures, and charts. Additionally, Unsiloed uses a specialized decoder that considers industry-specific terminology, ensuring accurate interpretation even when terms have different meanings across various business contexts.
To validate its claims, Unsiloed conducted a test involving 542 financial documents supplied by a client. The software achieved 99.78% accuracy across roughly 84,000 extracted data points, compared to major AI models like Microsoft's Azure Document Intelligence and the client's in-house extraction tool. Unsiloed also outperformed competing AI document-processing tools on olmOCR-Bench, a benchmark developed by Allen AI, scoring 88.0 compared to 66.0 for Reducto and 39.9 for Unstructured.
The benchmark assessed the systems' performance on 1,403 PDF pages, evaluating their ability to handle tables, mathematical expressions, scanned documents, and complex layouts.
With a current team of seven people, Unsiloed plans to use its latest funding round to hire more researchers and engineers, bolster its sales efforts, and develop proprietary datasets to enhance its technology further.
Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.