Pioneering legal search startup, DeepJudge, has launched a suite of tools for AI agents, which will enable firms to ‘build AI-powered workflows tailored to their needs without requiring them to restructure or upload their data’. (See AL Interview with CEO Paulina Grnarova, below, which explains in more detail what this is all about).
They have also announced US law firm Gunderson Dettmer and CMS Switzerland as new customers.
The company, which is based in Zurich and was founded by former Google engineers, specialises in finding relevant firm data, then surfacing it for lawyers to use in their workflows. Last summer they raised $10.7m in a Seed round.
They stated that: ‘Unlike traditional search, DeepJudge’s platform understands content, context, relevance, and institutional knowledge, allowing legal professionals to instantly find and leverage their firm’s collective expertise.’
As to the AI agents, they explained that: ‘The suite includes workflows that leverage full access to prior work product, such as determining whether opposing counsel has previously agreed to a specific clause. [There is also an] AI Workflow Builder, a low-code / no-code interface that allows firms to create custom AI-powered applications, multi-step workflows, and implement LLM agents.’ (Again, see interview below for more on defining AI agents.)
Grnarova, CEO of DeepJudge, commented: ‘AI should work for you – not make you work for it. Too many vendors require firms to ‘clean up their data’ before AI delivers value. That’s not solving the problem, it’s shifting the burden. But legal teams don’t work with perfectly structured information, and AI shouldn’t require it either.
‘Like a human, AI must navigate fragmented, inherited, and unstructured data without adding extra work. DeepJudge rethinks retrieval from the ground up to handle real-world legal complexity.’
While Yannic Kilcher, CTO of DeepJudge, said: ‘Legal data is trapped in silos: legacy databases, document management systems, SaaS platforms, and collaboration tools.
‘Extracting and indexing this data, while maintaining security at scale, is the critical challenge of making legal AI work. A search engine that does this doubles as both your critical tool for empowering your teams, and also as your intelligence layer for building, orchestrating apps and agents.’
And that’s what DeepJudge aims to provide.
And here’s a word from Joe Green, Chief Innovation Officer at Gunderson Dettmer: ‘What excites me most is how DeepJudge search powers our AI strategy….we can connect LLMs and AI agents to everything we’ve ever worked on – unlocking entirely new possibilities.’
—
AL Interview With DeepJudge CEO Paulina Grnarova
– First, congrats. How has customer engagement been since you started?
Thanks! Our customer engagement has been phenomenal. On the end-user side, we consistently see over 90% of attorneys as regularly active users, with thousands of lawyers already using our platform.
The reception among core roles, such as CIOs, has been equally strong. Our unique approach – building the workflows platform on top of our existing enterprise search engine to provide seamless, compliant access to internal data at scale – resonates deeply with their strategic thinking.
Moreover, our workflows are fully transparent and customizable, enabling firms to deliver unique, targeted GenAI experiences to their teams. This flexibility is crucial for aligning AI usage with a firm’s existing processes. As a result, many of our customers who had early access to AI Workflows have shared that their intent is to build their entire AI strategy on top of our search and workflows platform.
– There is a lot of blurriness when it comes to defining AI agents. In your case, what do AI Agents mean, and what are they doing that other aspects of standard programs +/or other direct inputs into AI are not doing?
The term “agent” has become somewhat blurry, often used as a marketing buzzword to mean different things depending on who’s selling the technology. In our experience, “agent” typically refers to LLM-based applications where the LLM itself serves as a decision-maker, or where multiple LLM requests build on each other. This is different from the more traditional LLM interactions, which typically involve a single prompt.
Our AI Workflows platform offers flexibility that can accommodate both. We provide the standard tablestakes GenAI features, such as Chat, Summarization, and Data Extraction, but also support more complex, multi-step agentic workflows that combine LLMs and Search in a highly synergistic way. Think – Deep Research, but using your own data. We also make sure everything is fully customizable, enabling firms to build their own unique, agent-driven workflows to meet their specific needs.
– Do you think lawyers are accepting AI agents? Do they want tasks to be linked together? Is loss of control an issue?
We are strong believers of “human in the loop”. Especially for complex agentic workflows, it is important that the human is in charge and informed about what’s going on under the hood. Our workflows are fully transparent—every prompt, every piece of data that is ingested, and every LLM output are clearly visible. We also show the search results that were considered in giving a response and link everything to the original source.
Given this transparency, we find human acceptance increases drastically, and the focus shifts away from fears about loss of control and towards AI agents assisting and augmenting humans in their most critical tasks.
– Where next? Given your successes, what next for the company?
We’re growing rapidly across all fronts, expanding our customer base in both Europe and the US while scaling our team across all functions.
The most exciting part is what’s next for our product. Once you have fast, efficient, and compliant access to a law firm’s internal knowledge at scale, the possibilities are big.
Our mission is to empower every individual in a law firm with access to their collective knowledge. We’re especially motivated when we see our customers scale AI across the entire firm, driving meaningful change and delivering strong ROI.
Thanks Paulina.
—
You can find more about DeepJudge here.