Ryan Samii — Why I joined Hebbia

As an M&A lawyer, I spent thousands of hours in front of my sacred dual monitors. Each screen developed its own identity.

My right screen was for “real” work, like surgically editing an agreement to favor my client’s position. This work was challenging and enjoyable.

My left screen was for supporting work: sifting through pages of diligence; dipping and diving through my firm’s document management system for relevant precedent; conducting needle-in-the-haystack searches in deal data rooms; poring over dense legalese to identify the few buried points of substance. This work served a purpose, but it was neither challenging nor enjoyable.

The unfortunate reality for lawyers, and tens of millions of other knowledge workers, is that this type of rote work – my ‘left screen’ work – comprises a disproportionate amount of the job.

What if this reality changes? What if the nature of knowledge work evolves? What would that look like and what would that enable?

I am excited to announce that I have joined Hebbia, an AI company that is asking and answering these questions.

Overpromise and Underdeliver, an AI Story

We are now over a year removed from headlines like “End of the Billable Hour?” (WSJ, May 2023), forecasting the rapid AI-driven revolution of legal services.

On the ground, however, a different reality has emerged. Across firms and sophisticated in-house legal teams, lawyers are hitting technical and functional limits with AI tools. With each new or unsolved obstacle, the trough of disillusionment is deepening.

Retrieval augmented generation (“RAG”), the go-to AI architecture for enterprise products, is failing. Sophisticated legal workflows – like analyzing an entire corpus of patents for prior art or comparing restrictive covenants across a suite of financing agreements – are proving too difficult for RAG. RAG struggles with the multi-step and data-heavy analyses that lawyers consistently require.

Chatbots, the go-to AI form factor, are failing. Thought exercise: before ChatGPT, what did a chatbot represent to you? For most, it was a tool for disputing charges on UberEats or processing a return on Nike’s website. Chatbots work for some use cases. Sophisticated knowledge work – which requires repeatability, traceability, and end-to-end workflows – is not among them.

Hebbia is Different

Hebbia is an intelligence tool that breaks down complex tasks into step-by-step actions carried out by AI agents.

Leading financial institutions (with $15 trillion+ in assets under management) and global law firms are using Hebbia’s ‘Matrix’ to (1) synthesize information from unstructured sources of data and (2) leverage the resulting structured data in familiar and novel ways.

From dissecting public company filings to analyzing complex merger agreements, Matrix engages in sophisticated step-by-step workflows that mirror how analysts, investors or lawyers approach their work. Matrix delivers on this promise for two key reasons:

• Technical Differentiation – Hebbia productionized RAG in 2020, putting the technology into the hands of thousands of knowledge workers. In the years that followed, a simultaneously painful and clarifying conclusion emerged… RAG does not cut it for serious work. Hebbia has developed a proprietary, patent-pending AI architecture that, unlike RAG, sources full documents without losing any context and decomposes user questions into multi-step agentic execution.
• Product Differentiation – Hebbia’s Matrix is a visual data grid that (1) decomposes a user’s query into step-by-step AI actions, (2) retrieves the appropriate data sources (e.g., side letters, credit agreements, depositions, or state regulations), thousands at a time if necessary, and (3) operates upon that data through a lawyer’s workflow of summarization, analysis, abstraction, comparison and/or synthesis – where separate AI agents own each step. This goes far beyond a chatbot.

Matrix is like Excel for words. Lawyers now have the ability to build or leverage automated workflows that eliminate drudgery and enable higher-order, strategic counseling.

What’s Next

Most technological shifts follow a similar journey. First, early novelty. Then, enduring utility.

Remember the ‘early novelty’ era of the internet? Once you got through access-via-dial-up, the ensuing product was just online versions of real-world analogs. Cookbook recipes, online. The New York Times, online. ESPN box scores, online.

It was novel! It was incredible! And yet, with the benefit of hindsight, it was so… limited. We were still a few technical and product-level breakthroughs away from the era of enduring and unpredictable utility – search engines, social media, and beyond.

Where do you think we are on the AI journey?

I joined Hebbia because I believe (1) the shift from early novelty to enduring utility is imminent and (2) Hebbia is, and will continue to be, a key force in this transition for the legal and financial industries.

I was captivated by CEO George Sivulka’s vision from our first conversation. We are setting out to shape the future of knowledge work. Investors have taken notice too, with $30M+ invested by the likes of Index Ventures, Radical Ventures, and Peter Thiel (whose first 5 personal pre-seed investments are Facebook, Palantir, DeepMind, OpenAI, and Hebbia).

Through my first couple months at Hebbia, my conviction has only deepened. This product is a gamechanger and this team is the strongest I have encountered.

If you are interested in learning more about how Hebbia is making a difference for law firms and in-house legal teams, please reach out (ryans@hebbia.ai) – and if you are interested in joining our growing team, we are hiring across the board!