Automating 90% of finance and legal work with agents

Investors, bankers, consultants, and lawyers spend countless hours combing through market and equity research, virtual data rooms, contracts, and regulatory filings to make high-stakes decisions.

Hebbia⁠(opens in a new window) set out to change that with Matrix, a multi-agent AI platform designed to handle the most complex financial and legal workflows end-to-end.

Rather than relying on a single AI model, Matrix orchestrates multiple AI agents in parallel, leveraging OpenAI’s o3‑mini, o1, and GPT‑4o all at once. The result: an “AI associate” that can perform in seconds what used to take entire teams days or weeks, and deep research that can process any amount of offline data to automate 90% of finance and legal work.

“We’re not just building a chatbot. We’re creating an agentic operating system that tackles the world’s most complex work.”

George Sivulka, CEO at Hebbia

Achieving state-of-the-art accuracy for professional tasks

Working with early clients, the Hebbia team recognized the key limitation in today’s AI-powered research isn’t the models themselves - it’s information retrieval over the world’s private information. 

While web search almost always retrieves answers from online sources, Retrieval-Augmented Generation (RAG)-based tools struggle for offline documents. Oftentimes, answers aren’t explicitly stated in documents, so traditional search falls short. 

Hebbia instead built a distributed orchestration engine that enhances accuracy for deep research tasks in finance and law.

The engine overcomes the limitations of RAG and effectively gives OpenAI’s models an “infinite” context window, creating the most accurate deep research agent for high value offline data. 

Hebbia with o1 achieves 92% accuracy—up from 68% with out-of-the-box RAG—on a rigorous benchmark spanning both quantitative and qualitative tasks across complex legal and financial documents.

Hebbia’s Matrix gives OpenAI models an infinite effective context window.