The impact of AI in the legal sector [Q&A]

AI is changing many industries. In the legal sector it's altering how businesses operate, automating routine tasks and boosting productivity for lawyers.

We spoke to Alon Shwartz, CTO and founder of Trellis AI, to find out more about AI’s transformative effect on the legal world.

BN: What are the core technical challenges in trial court litigation workflows?

AS: The fundamental bottleneck in civil litigation is data processing efficiency. Legal professionals currently operate with highly manual workflows for document analysis and knowledge extraction, creating a scaling problem as case complexity increases. A single motion brief might require processing thousands of documents to identify relevant precedents and bracket useful arguments, making the process computationally inefficient and error-prone. While this isn't necessarily a new challenge, it limits the number of cases attorneys can take on.

BN: What role does AI play in overcoming these challenges?

AS: ML systems excel at large-scale pattern recognition and information extraction tasks. The key technical advantage is parallel processing capability—being able to analyze millions of documents simultaneously rather than sequentially. This shifts the computational complexity from O(n) to O(1) for many core legal research tasks. The ML pipeline can handle document ingestion, semantic analysis, argument extraction, and pattern matching across the entire corpus of legal documents. Critical questions that traditionally created uncertainty in legal research -- Did I read all the relevant cases? Was there an angle I missed? How persuasive is my argument? -- can now be answered with heightened confidence due to the system's comprehensive analysis capabilities. Litigation attorneys can effectively see the entire legal landscape in a single view.

Our LLM models are specifically trained on legal terminology and based on state trial court data, not appellate data, enabling them to understand complex legal arguments and identify subtle patterns in cases that might be missed in manual review.

BN: What's the technical architecture behind Trellis AI's approach to litigation workflow automation?

AS: The key technical differentiator is our focus on state trial court data. While other AI solutions primarily leverage appellate case law, our comprehensive database of state trial court records provides a richer training dataset that maps more closely to day-to-day litigation patterns.

Until recently, legal research technology was primarily focused on appellate court data. However, the real challenge lies in processing the millions of state trial court records and documents produced daily. Finding relevant cases with matching causes of action and similar fact patterns traditionally required hours or days of manual search -- effectively a needle-in-a-haystack problem at scale.

The platform addresses this through a pipeline of specialized models that automate compute-intensive workflows:

• Motion drafting: Uses multi-step query to train focused models on millions of successful motions
• Multi-document timeline generation: Leverages multiple models to digest various documents and generate a graph timeline of key dates, events, and sources
• Case similarity analysis: Employs advanced embedding techniques to identify relevant precedents

The core innovation is making this massive volume of daily trial court records computationally tractable. We've developed custom models designed to perform different types of analyses on legal documents. By efficiently processing and extracting insights from these complex, unstructured materials, these models render stacks and stacks of state trial court data usable at unprecedented scales.

BN: How does this technological approach impact legal practice?

AS: The system fundamentally changes the information processing pipeline for legal work. Instead of manual document review and analysis being the bottleneck, attorneys can leverage automation to handle the computational aspects of legal research and writing.

Consider the case filing workflow. Traditional approaches require sequential human processing -- read complaint, research similar cases, analyze judge history, develop strategy, taking into account the jurisdiction, judge, opposing counsel, as well as similar cases. Our system parallelizes this entire pipeline. As soon as a case is filed, our models simultaneously:

• Extract key case attributes using named entity recognition
• Identify similar historical cases through multi-step matching
• Analyze judge behavior patterns using historical data
• Generate preliminary strategy recommendations based on outcome prediction
• Draft initial case assessment documents by matching similar cases in our massive trial court database

This shifts the attorney's role from information processing to high-level strategy and client consultation. The system handles the complex computational tasks while the attorney focuses on legal reasoning and client advocacy.

The technical architecture supports real-time updates and continuous learning. As new cases are filed and resolved, the system automatically updates its knowledge base and refines its models. This creates a feedback loop that continuously improves accuracy and relevance.

From an engineering perspective, the goal is to transform litigation from a primarily manual process into an automated, data-driven workflow. This allows legal teams to operate with greater efficiency and make more informed strategic decisions based on comprehensive data analysis. The system's scalable architecture can handle increasing volumes of legal data while maintaining consistent performance and accuracy.

The end result is a significant reduction in the computational burdens of legal research and analysis carried by attorneys, allowing them to focus on the strategic, relational, and advisory aspects of their legal practice -- the tasks that require human judgment and expertise.

BN: What are the biggest challenges facing trial court litigators today and what role does AI play in overcoming them?

AS: Civil litigation is flooded with busy, time-consuming work. Many litigation attorneys spend too much of their time doing low-value work like processing -- rather than practicing -- the law. They might spend hours, sometimes days, searching for and reading similar case briefs, extracting a large number of potentially relevant and useful arguments, in order to draft a 'winning' motion or answer. While this isn't necessarily a new challenge, it is limiting the number of cases attorneys can take on.

AI is great at doing one thing that humans cannot do. It can process enormous amounts of information in a matter of seconds. It can then detect patterns, recognizing trends that a human would never be able to. This information processing and analysis is the busy, low-value work mentioned above. It includes the mundane and tiresome tasks associated with legal research -- the copying and pasting of snippets of texts, the organizing and outlining, the summarizing of documents. AI can do this really well, and it can do it in a way that mitigates the uncertainties of the law.

Questions like, did I read all of the relevant cases? was there an angle I missed? how persuasive is my argument? can all be answered with a heightened sense of confidence because the processing power of AI garners an entirely new view of the litigation landscape. That is to say, litigation attorneys can pretty much see anything and everything with a single glance.

BN: How does Trellis AI empower legal teams to handle the increasing complexity of litigation?

AS: Trellis AI is a legal productivity platform. It's designed specifically for the complex needs of trial court litigation, automating some of the most cumbersome and time-consuming tasks, like motion drafting, case assessment, multi-document timeline generation. It is powered by Trellis, the most comprehensive database of state trial court data in the country, which also makes it the most credible.

Other AI solutions leverage appellate cases or federal data, which makes their AI solution less applicable to the needs of state trial court litigation. This is what we are pointing to when we talk about the increasing complexity of litigation. Until recently, legal research meant scouring case law from the appellate courts. But now, litigation attorneys have access to the millions of state trial court records produced every single day. The effort of finding relevant cases that share my causes of action, and within those, identifying the motions that share the same facts as my case, takes hours and sometimes days. This effort makes the entire process feel like finding a needle in a haystack. Trellis AI is designed to resolve this tension, rendering the stacks and stacks of state trial court documents usable in a scalable manner.

BN: What long-term impact do you envision AI having on client advocacy and the way attorneys deliver value to their clients?

AS: The status quo of civil litigation leaves attorneys with very little time to actually litigate, to think strategically about a case and its effective resolution. Most of their time is spent on research and writing tasks.

For example, the 'industry standard' now is for an attorney to receive an alert when one of their clients is sued. The alert may include one or two lines, a high-level summary of the complaint, but nothing more substantive than that. The attorney needs to read the complaint and create some strategy before calling the client. And they need to do this quickly—before a competing firm has a chance to call the client. This leaves little time to strategize, which makes the call superficial and unproductive.

The purpose of a product like Trellis AI is to reimagine the attorney-client relationship. What if, instead of a one or two line summary, the alert system automatically generated a fully drafted case assessment, almost a roadmap for how to deal with this case? How might that phone call look different? What if an attorney could receive an alert, and then immediately call their client and say, "Hey, you're being sued. These are the strengths and the weaknesses of the case. I have already drafted three potential litigation strategies, which take into account the judge on the case, the witnesses in the action, and the potential jury pool."

The client could leave that conversation with a clear understanding of the cost, complexity, and risks of the legal action. They could feel confident knowing their attorney is not only prepared but already steps ahead.

This is the promise of Trellis AI: transforming litigation from a reactive process into a proactive, strategic partnership. Attorneys can move faster, act smarter, and deliver exceptional value -- redefining what it means to be a trusted legal advisor in today's competitive landscape.

Image credit: phonlamai/depositphotos.com