We didn’t see this coming two years ago. Not because the technology was unproven, but because we assumed the market would reward the platform with the broadest moat, not the one with the deepest niche.
Rogo just announced it tripled its annual recurring revenue to over $50 million, vaulting past Hebbia in the race to become Wall Street’s default AI operating system. [[51]] The startup, founded by former Lazard banker Gabriel Stengel, went from roughly $2 million in ARR in 2024 to over $15 million in 2025, then leaped past $50 million in 2026. [[52]] Hebbia, by contrast, reported $13 million ARR as of mid-2024 and has since claimed profitability but has not disclosed updated top-line figures. [[21]] The gap is now wide enough that the question is no longer who is ahead, but whether the two companies are even playing the same game.
Let me set the stage. Both companies emerged around 2020-2022 with a shared conviction: that generic AI tools fail in high finance because they don’t understand the language of deals, the structure of pitch books, or the stakes of a mis-cited number. But they diverged sharply on what to build.
Hebbia, founded by Stanford PhD student George Sivulka, built Matrix, a multi-agent AI platform designed to ingest thousands of unstructured documents — PDFs, transcripts, scanned images — and answer complex, multi-step queries with sentence-level citations. [[43]] Its proprietary ISD (Inference, Search, Decomposition) architecture rejects standard Retrieval-Augmented Generation in favor of an "infinite" context window that decomposes questions into agentic sub-tasks. [[41]] Hebbia’s insight was that for a financial analyst or a lawyer, an un-sourced answer is worthless. By 2024, Matrix was used by 30% of the top 50 global asset managers, including BlackRock, KKR, and Carlyle. [[28]] The company raised $130 million in Series B funding in July 2024 at a roughly $700 million valuation, led by Andreessen Horowitz. [[21]]
Rogo took a different path. Rather than building a universal document reasoning engine, Stengel and his co-founders focused on one thing: the transactional workflow of investment banking. Rogo’s AI generates the specific deliverables banks produce for clients — financial models, pitch decks, IPO documents, memos. [[51]] Its platform is not a general-purpose query tool; it is a purpose-built factory for sell-side outputs. The company fine-tuned OpenAI’s models on financial datasets from S&P Global, Crunchbase, and FactSet, and trained its agents to understand the context of deal preparation rather than treating every document as a generic text corpus. [[55]] By December 2024, Rogo had grown ARR 27x and was saving analysts more than 10 hours per week on meeting prep and company profiling. [[55]]
Here is where the analysis gets interesting. The surface narrative says Rogo won because it raised more money — $310 million total across five rounds versus Hebbia’s $161 million — and scaled faster. [[54]] But that framing misses the deeper structural difference.
Rogo’s ARR explosion from $2 million to $50 million in two years was not just a function of capital. It was a function of workflow lock-in. Rogo doesn’t just answer questions; it produces the artifacts that bankers get paid for. When a managing director at Moelis or Lazard uses Rogo to generate a client-ready pitch book, the tool becomes embedded in the firm’s production pipeline, not just its research process. [[51]] The switching cost is enormous because the output is not a summary — it is the deliverable itself.
Hebbia, by contrast, sits upstream. Its Matrix platform excels at deep research and document intelligence: analyzing thousands of deal memos, extracting risk factors from SEC filings, synthesizing management call transcripts. [[76]] But it stops short of producing the final work product. As one industry comparison put it in mid-2026: "Rogo is finance-specific by design; its tooling is built around the way deal teams actually work, with capabilities tuned for investment-banking and sell-side tasks such as building materials, running comparable analyses and preparing for transactions." [[72]] Hebbia solves the reading problem; Rogo solves the doing problem.
This distinction matters because it maps directly onto the data access model. Rogo brings its own data through partnerships with LSEG, FactSet, and S&P Global, meaning a banker can query market intelligence without leaving the platform. [[79]] Hebbia processes your proprietary documents — your data room files, your internal memos, your management presentations. [[79]] One is a walled garden with premium content; the other is a tool for your private kingdom. The right choice depends on whether your bottleneck is finding external data or making sense of your own.
Now the contrarian piece, and the part that keeps me up at night as someone who has seen too many "unbeatable" leads evaporate.
Rogo’s $2 billion valuation on $50 million ARR implies a roughly 40x revenue multiple. [[51]] In a bear market where enterprise software multiples have compressed, that is a bet on continued hypergrowth. The company would need to reach roughly $200 million ARR within three years to justify that valuation under conventional SaaS multiples. That is not impossible — Rogo’s trajectory from $2 million to $50 million in two years suggests the underlying demand is real — but it is not guaranteed.
The bigger risk is not market saturation. It is commoditization from above. Rogo’s architecture relies on fine-tuning frontier models like GPT-5 and Gemini 3. [[53]] If OpenAI or Anthropic decide to build their own financial workflow layer — and they have the talent and capital to do so — Rogo’s fine-tuning moat could narrow quickly. Hebbia faces the same existential threat, but its proprietary ISD architecture and document-agnostic positioning give it a different kind of optionality: it can expand into legal, pharmaceutical, and other document-intensive verticals where Rogo’s finance-specific design does not transfer. [[24]]
There is also the question of customer concentration. Rogo serves 150 to 250 institutions, with 25,000 daily users. [[51]] In enterprise software, a handful of top clients can drive 40-60% of ARR. If one of Rogo’s anchor banks — say, Lazard or Jefferies — decides to build in-house or switches to a competitor, the revenue impact would be immediate and material. The company does not disclose net revenue retention, which is the single metric I would ask for before making any judgment on long-term unit economics.
Hebbia, meanwhile, is not standing still. In August 2026, the company relaunched Matrix with features that more closely mirror Rogo’s workflow automation tools. [[75]] The move signals that Hebbia recognizes the need to move upstream from pure research into production. But playing catch-up on product features is a dangerous game when your rival has a 4x ARR head start and a $2 billion valuation to fuel sales expansion.
Innovation without integrity is just noise. That sentence has guided my coverage of this space for the better part of a decade, and it applies here with unusual force.
What Rogo and Hebbia are building matters beyond the quarterly numbers. Financial services is an industry where information asymmetry — who has the data, who can process it fastest — directly determines who extracts value. For the past fifty years, that asymmetry has been maintained by armies of junior analysts working 100-hour weeks, building models and assembling decks at the expense of their health and their curiosity. [[66]] AI does not eliminate the asymmetry; it shifts who holds it. The question is whether the shift concentrates power in fewer hands or distributes it more evenly.
Rogo’s CEO, Gabriel Stengel, has been explicit that his goal is not to replace junior bankers but to free them for higher-value work. [[10]] The company’s AI agent, Felix, automates the most mechanical tasks — pulling data, formatting slides, running comparable analyses — so that analysts can spend more time on client relationships and strategic thinking. [[9]] That is the optimistic scenario. The pessimistic one is that firms use these tools to thin their junior ranks, maintaining output with fewer people, and further concentrating decision-making authority among senior partners who are already over-leveraged.
The evidence so far is mixed. Rogo says its tools save analysts 10+ hours per week, but it has also partnered with firms like Rothschild & Co and Jefferies that are known for lean deal teams. [[2]] Hebbia claims its platform is a "thought partner for senior decision makers," but the reality is that any tool that automates due diligence at scale reduces the demand for manual review. [[10]]
We didn’t build this technology to replicate the old hierarchy. But the old hierarchy is the one writing the checks.
For builders and allocators reading this, the actionable question is not "Rogo or Hebbia?" It is "what kind of intelligence are you trying to build?"
If you believe the most valuable financial AI will be the one that produces the final output — the model, the deck, the memo — then Rogo’s workflow-centric approach has the stronger near-term business case. Its ARR trajectory, its blue-chip investor base (Sequoia, Thrive Capital, J.P. Morgan, Kleiner Perkins), and its deepening partnerships with LSEG and FactSet all point toward a company that has found product-market fit in a specific, high-value niche. [[2]] [[53]] The risk is that this fit is too narrow — that the $50 million in ARR represents the low-hanging fruit of investment banking automation, and the next $100 million requires entering adjacent verticals where Rogo has no advantage.
If you believe the endgame is a universal document reasoning layer that spans finance, law, and regulated knowledge work, then Hebbia’s architecture is the more defensible bet. Its ISD architecture is patent-pending and demonstrably more accurate than standard RAG — 92% accuracy versus 68% on a recent benchmark. [[41]] Its customer base includes top-10 global asset managers and the U.S. Air Force. [[34]] The challenge is commercial: Hebbia needs to grow ARR from $13 million to $50 million-plus without the luxury of a finance-specific sales motion. The relaunch of Matrix in August 2026 suggests management understands this urgency. [[75]]
The market will not wait for either company to prove its long-term thesis. The generative AI market in banking, financial services, and insurance is projected to grow from $1.9 billion in 2025 to $18.5 billion by 2034. [[3]] That is a 10x expansion in a decade, and it will attract every well-capitalized player in enterprise software.
Open source is a handshake, not a contract. The same is true of early ARR leads. Rogo has the stronger handshake today. But the contract — the question of who builds the trusted infrastructure for financial intelligence — is still being written.
And if there is one thing I have learned from years in this industry, it is that the company that wins on trust wins the long game. Rogo has the revenue. Hebbia has the architecture. The next 18 months will tell us which of those is harder to replicate.
