How AI Is Being Used to Clear Court Backlogs in LA
Courts across the globe are buckling under the weight of ever‑growing caseloads, and Los Angeles is no exception. In response, the Los Angeles Superior Court has begun piloting an artificial intelligence system designed not to replace judges, but to give them breathing room by cutting down on repetitive, time‑consuming tasks.
At the center of this experiment is an AI tool called Learned Hand. Rather than issuing decisions on its own, the software works in the background of civil cases: it digests lengthy filings, structures the evidence, and drafts preliminary versions of rulings for judges to review and edit. The aim, according to Learned Hand founder and CEO Shlomo Klapper, is straightforward-free judges from administrative overload so they can devote more of their time to the complex legal reasoning that only humans can provide.
Klapper describes the situation bluntly. Courts are “under tremendous strain,” he said in an interview, with dockets growing faster than the judicial system can absorb. Caseloads keep climbing, but additional staff and new judges are not arriving at the same pace. At the same time, advances in generative AI are dramatically lowering the cost of producing legal documents, making it far easier and cheaper for litigants and lawyers to generate motions, briefs, and other filings.
That technological shift has a paradoxical effect. On one hand, AI helps lawyers and self‑represented litigants prepare documents more quickly. On the other, it increases the volume of materials judges must read. Klapper notes that filings in one context rose by 49%, from 4,100 to 6,400, underscoring how rapidly the workload can swell when tools make document creation almost frictionless. Rather than resisting that trend, Learned Hand is intended to balance it by giving courts their own AI‑driven assist.
In practice, the tool starts by reading and summarizing the parties’ submissions. Instead of forcing a judge or clerk to comb through dozens or even hundreds of pages to understand the dispute, the AI produces concise overviews of the issues, arguments, and requested relief. These summaries can be customized to highlight specific questions-such as jurisdiction, evidentiary disputes, or procedural defects-reducing the time spent merely getting oriented in a case.
Learned Hand also helps organize and tag evidence. Civil cases can involve sprawling records: contracts, email threads, expert reports, financial spreadsheets, and more. The AI system can categorize these materials by relevance, topic, or procedural stage, making it easier for court staff to retrieve exactly what they need at the moment it matters. That kind of structured organization is especially valuable in busy courts where multiple complex cases are moving forward at once.
Another key feature is the generation of draft rulings. The AI can produce an initial framework for an order or opinion based on the filings and relevant legal standards. Crucially, this is not a final decision: judges retain full control over the reasoning, the outcome, and the language. The draft is meant to serve as a starting point, offering a suggested structure and possible analysis that a judge can accept, modify, or reject entirely.
Supporters of the program emphasize that judicial discretion remains central. The AI is framed as a smart assistant, not an automated judge. It does not decide who wins or loses, and it does not issue binding orders. Instead, it handles the kind of repetitive, pattern‑based tasks-summarizing, sorting, and drafting boilerplate sections-that consume countless hours of court time but do not themselves require uniquely human judgment.
The stakes are high. In an overloaded system, delays can stretch on for months or even years, eroding public confidence and raising costs for everyone involved. Litigants wait longer for closure, lawyers must manage clients’ frustration, and judges are forced to triage their attention. If AI can even modestly speed up the handling of routine motions or preliminary issues, it could translate into significantly shorter queues and quicker access to justice.
That said, the experiment raises important questions about transparency and fairness. One concern is whether parties should be told when an AI system helped produce the draft of a ruling. Another is how to ensure that the AI’s suggestions are free from hidden biases embedded in the data it was trained on. Courts must also be able to audit and explain their decisions, which becomes more complex when software is involved anywhere in the process, even at the drafting stage.
There is also the issue of quality control. AI systems can misinterpret nuances, overlook context, or hallucinate facts or citations if not carefully constrained. For that reason, the Los Angeles pilot is structured so that judges and clerks remain the ultimate gatekeepers. They are expected to verify the AI’s work, correct errors, and ensure that the final decision reflects the law and the record-not the blind output of an algorithm.
Beyond immediate efficiency gains, initiatives like this could reshape how court staff are trained and deployed. Clerks might spend less time on rote summarization and more on substantive research. Administrative staff could be reoriented toward managing and validating AI‑generated work product. Over time, standard workflows for handling civil cases may evolve to assume an AI assistant is always present in the background.
There are potential benefits for self‑represented litigants as well. When judges and court staff can process matters more swiftly, people without expensive legal counsel may face fewer delays and clearer guidance. Faster turnaround on routine rulings-such as scheduling orders or straightforward motions-can make the entire process less intimidating and more predictable for those unfamiliar with the legal system.
The Los Angeles pilot will also provide valuable data on where AI tools actually save time and where they do not. For example, the technology might prove highly effective at handling standard pre‑trial motions but less helpful in fact‑intensive disputes or novel legal questions. Those insights will inform whether and how the program should be expanded, refined, or limited to particular types of cases.
If the experiment succeeds, it could serve as a model for other jurisdictions facing similar bottlenecks. Many courts are already exploring digital case management, electronic filing, and remote hearings; AI‑assisted drafting and evidence organization may become the next step in that modernization process. However, any broader rollout will need to address funding, training, cybersecurity, and strict safeguards for confidential information.
Ultimately, the Los Angeles Superior Court’s collaboration with Learned Hand underscores a broader shift in the legal world. AI is no longer just a tool used by law firms and litigants; it is starting to enter the heart of judicial administration itself. The core challenge is to harness the speed and scale of these technologies without compromising the independence, integrity, and humanity of the courts.
For now, the pilot remains a test of a simple but ambitious hypothesis: that by offloading repetitive tasks to a carefully designed AI system, judges can reclaim time for the kind of nuanced, principled decision‑making that cannot be automated. As filings continue to increase-driven in part by the same AI advances that enable this new tool-the results in Los Angeles may offer an early glimpse of how justice systems worldwide adapt to the age of artificial intelligence.