Hebbia ships new version of Matrix as AI research tools crowd Wall Street

Hebbia has released an updated version of Matrix, its document-analysis product, as the company works to defend an early lead in AI tools built for Wall Street research desks, according to Business Insider.

The original Matrix product took a specific bet on interface. Rather than a chatbot answering one question at a time, users upload a set of documents and pose several questions at once. The system returns a grid: one row per document, one column per question. That format matched how analysts already worked in Excel and made the output easier to audit than a conversational thread.

In a blog post accompanying the release, CEO George Sivulka argued that AI capability alone does not create value, and that the product layer is what turns a model into something a professional can use. He compared the relationship to fire and the torch, or the wheel and the chariot, per Business Insider's account of the post.

Hebbia was founded in 2020. In 2024, it raised USD 130M in a round led by Andreessen Horowitz. That round put the company among the better-capitalized entrants in the AI-for-finance category, alongside AlphaSense, Rogo, Linq, and a growing set of horizontal players extending into the same workflow.

The competitive backdrop is what makes the Matrix update notable. When Hebbia launched, the grid interface was differentiated. Four years later, most large vendors serving buy-side and sell-side research have shipped some version of multi-document Q&A, whether as a standalone product, a feature inside a broader terminal, or an agentic workflow built on top of a foundation model. The moat is no longer the idea of asking many questions of many documents. It is the accuracy of retrieval, the latency of the grid, the auditability of each cell, and the depth of integration into existing research stacks.

Business Insider's piece is paywalled beyond the setup, so the specific feature set of the new Matrix version is not public in the excerpt available. What is public is the framing: Sivulka is positioning Hebbia as the product layer for AI in finance, not as a wrapper on someone else's model.

The broader signal is that the AI-for-finance category has moved past the phase where a single interface choice was enough to define a company. Product-layer arguments now have to be defended feature by feature, benchmark by benchmark, and desk by desk.