Researchers from ETH Zurich, the New Economic School, and Imperial College London conducted a large-scale field experiment with Pakistan's judiciary, covering 1,559 judges across 118 courts. The tool, JudgeGPT, an AI assistant built on OpenAI's GPT-4, uses retrieval augmented generation to search a database of 129,235 documents, including 128,292 court rulings and 943 Pakistani laws. Judges were split into three groups: one received JudgeGPT access plus targeted training (six 90-minute lectures over three weeks), a second got AI access with a general seminar, and a control group attended the seminar without access. After 40 weeks, trained judges averaged nearly 60 logins and over 200 prompts, while the comparison group averaged about 20 logins and fewer than 50 prompts. Districts with more trained judges resolved about 1,848 extra cases per year per district, a 6.3 percent bump. The researchers estimate savings of about $38.50 per dollar invested, with conservative estimates at "at least" $10 per dollar. Ruling quality held steady or improved, with no evidence of increased gender or religious bias.