Agentic AI that turns unstructured documents into structured answers for finance and legal teams
HQ: New York, NY | Founded: 2020 | Employees: ~200 (2026, est.) | Stage: Late-stage VC (Series B) | Website: https://www.hebbia.com
Hebbia was founded in August 2020 by George Sivulka, then a Stanford economics PhD student and ex-bridge-investor, who set out to fix the fundamental problem that documents don't answer questions — people have to read them. Its initial pitch made the search-like capability of 'Ctrl+F for anything' feel overambitious; backers at the seed stage included Peter Thiel, Floodgate, and Naval Ravikant.
Hebbia's breakthrough product is Matrix, launched in 2022: an agentic research platform that ingests tens of thousands of documents and produces a spreadsheet-like grid of structured answers ('bento' cells) where every answer is decomposed into its underlying sources with traceable citations. The proprietary architecture, Iterative Source Decomposition (ISD), parallelizes document understanding and lets analysts validate each cell against the raw evidence — critical for due diligence, covenant monitoring, and board work in finance and law.
The company raised a ~$30M Series A led by Index Ventures in September 2022 and a $130M Series B led by Andreessen Horowitz in July 2024 (with Index, GV, and Peter Thiel participating, plus angels Eric Schmidt and Jerry Yang) at a ~$700M valuation on ~$13M of profitable revenue. Reported ARR grew to an estimated ~$50M by September 2026. Named customers include BlackRock, KKR, Carlyle, Centerview Partners, MetLife, U.S. Air Force, Gunderson Dettmer, Fenwick, Orrick, and TowerBrook, and Hebbia claims to serve a large share of the world's biggest asset managers and over $15T in managed assets. In 2025 it acquired FlashDocs (generative slide-deck and artifact creation) and integrated with Microsoft Azure AI Foundry for GPT-5; in 2026 it launched the Max deal-team collaborator and Matrix 2.0, and took a strategic investment from the Presight–Shorooq Fund (G42-linked), with an SF and London office build-out. Hebbia is SOC 2 Type II and ISO/IEC 42001 certified, and states it does not train on customer data.
Hebbia is the quiet winner of the enterprise-document AI race for finance and legal: it is profitable, model-agnostic, and deeply embedded in the world's biggest asset managers and law firms, with an estimated ~$50M ARR. Founder George Sivulka built Matrix around a genuinely different interaction model — a spreadsheet-like grid where AI decomposes unstructured documents into citable, structured answers rather than chat summaries — which gives it an edge over both generic LLM chatbots and search tools. The bull case is that knowledge-work professionals will keep paying for source-grounded, auditable answers and Hebbia keeps accruing the best logos while staying model-agnostic. The bear case is platform competition: Microsoft (its own Azure AI Foundry partner and now a competitor), OpenAI (which owns the models Hebbia routes on), and well-funded rivals like Glean and Rogo could commoditize or out-flank it.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki