Hebbia
Agentic research at document scaleEnterprise AI research platform whose Matrix product runs agentic analysis across thousands of documents, returning structured answers with cell-level citations.
- Maker
- Hebbia
- Founded
- 2020
- Pricing
- No free tier · from Custom
- Last checked
- October 2026
Verdict
Enterprise AI research platform whose Matrix product runs agentic analysis across thousands of documents, returning structured answers with cell-level citations.
Score breakdown
vs Consensus, the category leaderInvestment banks, private equity firms, asset managers and large law firms running document-heavy diligence at scale
Key features
- Matrix — spreadsheet-style grid that runs questions across entire document sets and returns structured, cited answers
- Iterative Source Decomposition (ISD) — Hebbia's retrieval architecture, positioned as a replacement for RAG rather than a variant
- Deeper Research multi-agent system with Orchestrator, Planning, Retrieval, Analysis, Distillation and Reasoning roles
- Skills — reusable, shareable analyses that can reference PowerPoint, Word and Excel files
- Generates editable Excel spreadsheets and branded multi-slide decks directly from a prompt
- Microsoft Excel plug-in with inline cell citations delivered as Office notes
- Data integrations: FactSet, S&P Capital IQ, PitchBook, Preqin, Third Bridge, Fitch Solutions, ICE, Snowflake, SharePoint, Box, S3
- Model-agnostic execution across OpenAI, Anthropic and Google models
Strengths
- +Matrix's grid interface — documents as rows, questions as columns — is a genuinely better fit for diligence work than chat, letting one prompt run across thousands of filings at once
- +Every cell links back to the exact source passage, so outputs are auditable rather than taken on trust
- +Deep, finance-native data integrations: FactSet, S&P Capital IQ, PitchBook, Preqin, Third Bridge, Fitch Solutions and ICE fixed-income pricing
- +Strong compliance posture for regulated buyers — ISO/IEC 42001, SOC 2 Type II, GDPR and CCPA, with AES-256 at rest
- +Adopted at serious institutions including Morgan Stanley, MetLife, Centerview Partners and Latham & Watkins
Weaknesses
- −No free tier, no trial and no self-serve signup — evaluation is entirely sales-gated
- −Expensive and opaque: Hebbia publishes no rate card, and third-party reporting puts seats in the $3,000-$15,000/year range with large deployments in the mid-to-high six figures
- −Narrow vertical focus on finance and legal document workflows; a poor fit for general enterprise knowledge search
- −In legal work it analyzes but does not draft — no native redlining, playbook enforcement or clause-level suggestions
- −Accuracy is publicly unbenchmarked; the citation-first design makes errors auditable but no independent evaluation exists
Overview
Overview
Hebbia is a New York-based enterprise AI company founded in August 2020 by George Sivulka, Lukas Schmit and Tim Lupo. It raised a $130M Series B in July 2024 led by Andreessen Horowitz, with GV, Index Ventures and Peter Thiel participating, at a valuation of roughly $700M — reported at the time on $13M of ARR and described as profitable. Total disclosed funding is approximately $161M.
The flagship product is Matrix, which abandons the chat interface for a grid: documents are rows, questions are columns, and every cell is an answer linked back to the exact passage that produced it. That structure suits the work Hebbia’s customers actually do — running one question across thousands of SEC filings, credit agreements, CIMs or data-room documents and comparing results side by side. Underneath sits Iterative Source Decomposition, which Hebbia positions as a replacement for retrieval-augmented generation, decomposing a request into sub-tasks and iteratively retrieving across sources rather than embedding-matching a single query.
Hebbia sells exclusively to enterprises: no free tier, no trial, no published pricing. Named customers include Morgan Stanley, MetLife, Centerview Partners, Oak Hill Advisors, New Mountain Capital and Latham & Watkins, and the company reported serving roughly 30% of the top 50 asset managers by AUM as of mid-2024. Third-party reporting puts seats somewhere between $3,000 and $15,000 per year, with large deployments reaching the mid-to-high six figures — figures Hebbia has never confirmed.
Key Benefits
- Scale that chat interfaces cannot match: One prompt fans out across an entire document corpus, with results laid out for direct comparison rather than buried in conversation.
- Auditability by construction: Cell-level citations mean an analyst can verify any answer against its source in one click — essential where output feeds an investment memo or legal opinion.
- Finance-native data: Integrations with FactSet, Capital IQ, PitchBook, Preqin and ICE bring market data alongside the documents, not in a separate tool.
- Deliverables, not just answers: Recent releases generate editable Excel models and branded slide decks directly from a prompt, and an Excel plug-in carries citations into the spreadsheet.
Use Cases
- Private equity and M&A diligence — Run standardized question sets across an entire data room, surfacing change-of-control clauses, customer concentration and covenant terms across hundreds of documents at once.
- Public markets research — Compare disclosures, risk factors and guidance across years of filings and earnings calls for a whole coverage universe in a single Matrix.
- Credit and fixed-income analysis — Extract covenant terms from credit agreements and pair them with ICE bond pricing, spreads and yield history inside the same workflow.
- Legal document review — Law firms use Matrix to analyze contract sets during diligence, though drafting, redlining and negotiation remain outside the product’s scope.
Pricing
Checked October 2026 · USD- Seat-based annual contracts, quoted through sales
- No public rate card; third-party reports suggest roughly $3,000-$15,000 per seat per year depending on tier
- Large finance and legal deployments reported in the mid-to-high six figures annually
- Typically a one-year minimum commitment
- Includes ISO/IEC 42001 and SOC 2 Type II compliance, SSO and enterprise data controls