Smar NetworkSmar.inSmargrid.comNewsInsuranceRoomsGamesMeasurementsCooking

SMAR Daily Industry News

Daily news across AI, technology, finance, insurance, healthcare, automotive, real estate, and government

Facebook Telegram
← Back to AI News

Unsiloed AI Secures $7.5M to Make Enterprise Docs AI-Ready

📅 2026-11-18 | 📰 TechCrunch AI
#unsiloed ai #document processing #ai agents #seed funding #enterprise ai

Unsiloed AI, a San Francisco startup, announced a $7.5 million seed round led by venture firm Felicis, with participation from Y Combinator, Orange Collective, Uncorrelated Ventures and angels Aarthi Ramamurthy and Beth Turner. The funding will support hiring of researchers and engineers, expand sales, and build proprietary datasets for its document-processing platform.

The company’s technology tackles a core obstacle for enterprises adopting AI agents: extracting reliable information from tables, charts, and intricate layouts that traditional large language models often misinterpret. Unsiloed’s approach uses a “dual-stream” vision model that processes text tokens and image tokens separately, combined with heat-map-based chunking to focus compute power on the most information-dense sections of a document. A specialized decoder then applies industry-specific terminology to improve accuracy across finance, legal and healthcare contexts.

In a technical brief, Unsiloed reported 99.78 % accuracy on 84,000 data points drawn from 542 financial documents supplied by an enterprise customer, outperforming Microsoft’s Azure Document Intelligence, the customer’s in-house tool, and major AI models. On the public olmOCR-Bench benchmark, the startup scored 88.0, compared with 66.0 for Reducto’s system and 39.9 for Unstructured, across 1,403 PDF pages featuring tables, mathematical expressions, scanned pages and complex layouts. These results have not been independently verified.

Unsiloed already counts a bank, a law firm, and voice-AI startup Bolna AI among its early adopters. The bank uses the platform to process internal documents, while the law firm runs the software on its own servers for contracts and case files. Bolna AI leverages Unsiloed to ingest knowledge bases for building voice agents. Pricing starts at one cent per page, a cost structure the CEO says is enabled by directing processing resources only to the most demanding parts of a document.

Unsiloed joins a crowded market that includes Reducto, Unstructured, and document-intelligence services from Microsoft and Amazon. Its differentiation lies in the dual-stream architecture and industry-aware decoding, which the founders claim reduce hallucination risk inherent in probabilistic large language models.

Why it matters

Enterprises that rely on accurate document extraction can reduce costly errors and accelerate AI-driven workflows.

FAQ

Who led Unsiloed AI’s seed funding round?
Silicon Valley venture firm Felicis led the $7.5 million round, with participation from Y Combinator, Orange Collective, Uncorrelated Ventures, and angels Aarthi Ramamurthy and Beth Turner.

What accuracy did Unsiloed achieve on the financial document test?
Unsiloed reported 99.78 % accuracy across roughly 84,000 extracted data points from 542 financial documents.

How does Unsiloed’s pricing model work?
Pricing starts at one cent per page, reflecting the company’s approach of allocating more compute power to the most complex parts of a document.

Which industries are early customers of Unsiloed?
Customers include a bank, a law firm, and voice-AI startup Bolna AI, with the startup also targeting finance, legal services, and healthcare.

Source: TechCrunch AI. See the original source for full details.