Magdalena Administrative Court

Colombia’s judiciary faces severe backlog and slow due process: a substantial share of cases were pending in 2023 (about half of all cases), the average judge’s caseload exceeded 800, and an administrative ruling can take about 10 years to resolve. Judge María Victoria Quiñones and her team at Magdalena’s administrative court were overwhelmed by repetitive work—writing hearing minutes, transcriptions, summaries, and draft judicial documents—that consumed hours and limited how many rulings they could issue.

Quiñones and her aides adopted Microsoft Copilot in Teams under a Colombia judiciary beta program (about 20 judges initially) to summarize virtual hearings, identify attendees, draft and polish judicial documents, support transcriptions, review grammar and wording, compare draft rulings against prior case material, and manage internal data. The court also uses related Microsoft tools such as Power Automate and SharePoint, with custom Copilot prompts, while judges remain required to review, verify, and disclose AI use under national judiciary rules.

Hearing minutes that previously took six hours were produced in a few minutes with Copilot-assisted drafting, then verified and polished by the team. Quiñones reports she can now issue up to 20 rulings in a week, up from four before adopting Copilot. Comparing a colleague’s draft ruling to prior rulings on a topic—work that once took a whole morning browsing files—now returns an answer in seconds using more than 10,000 digitized internal cases.

• Generative AI in courts delivers the most value on repetitive drafting and summarization when humans verify outputs and AI is barred from deciding evidence or final legal judgments.
• Pairing Copilot with virtual hearings and existing case archives multiplies productivity beyond either tool alone.
• Training judges and staff, plus clear disclosure and data-security requirements, are central to judiciary AI adoption.