Colombia's Constitutional Court uses an NLP tool, PretorIA, to identify and cluster priority cases among 600,000+ annual constitutional-rights filings, cutting case-selection time from 96 days to about two minutes; no independent accuracy figure has been published.
96 days to ~2 minutes
Case-selection time (since 2020)
~4,500
Cases processed daily
600,000+
Annual tutela filings received
Details
Maturity
Established
Promoter
Constitutional Court of Colombia (Corte Constitucional)
Period
2018–2026
Keywords
judicial case management, natural language processing, docket triage, courts, case selection, legal tech
Context
Colombia's Constitutional Court receives more than 600,000 tutela (constitutional-rights protection) cases a year, a 1,900% increase over 20 years. To help identify cases raising important or recurring constitutional issues, the Court developed PretorIA between 2018 and 2020 with Universidad del Rosario and the Instituto Latinoamericano de Inteligencia Artificial (UBA IALAB, Argentina). Operational since 2020, the natural-language-processing system processes approximately 4,500 cases daily as a trained search engine that surfaces candidate cases for human judges, who retain sole decision-making authority. The tool was originally conceived with more automated decision-support functions but was deliberately scaled back to a case-identification and search-assistance role after concerns about the risks of a more autonomous version.
Results
Oxford's Institute for Technology and Justice cites a reduction in case-selection time from 96 days to roughly two minutes as a result of the tool. Unlike its Argentine sister tool PROMETEA, which claims 96% accuracy, no independently verified accuracy or error-rate figure for PretorIA itself has been published, so the efficiency gain should be read as a workflow-speed claim rather than proof that case-selection quality matches the prior manual process. Colombian digital-rights and human-rights organisations Derechos Digitales and Dejusticia have both published independent analyses examining the system's design and implementation challenges, including the practical difficulty of digitising a large volume of paper-based case files.
Implementation
Indicative cost
Medium (€50k–€500k) — No budget figure is published; classified as medium cost given a two-year multi-institutional development effort (2018-2020) with two academic partners.
Time to results
Medium (1–3 years) — Developed over roughly two years (2018-2020) before becoming operational in 2020.
Staffing & skills
Human judges (retain sole decision-making authority), Universidad del Rosario researchers, Instituto Latinoamericano de Inteligencia Artificial (UBA IALAB, Argentina)
Conditions for success
Human-in-the-loop design: system surfaces candidates only, judges decide
Cross-institutional academic partnership for development
Deliberate scope reduction from an automated decision-support design to a search-assistance role after risk concerns
Common failure modes
No independently verified accuracy or error-rate figure has been published for PretorIA itself
Digitising a large volume of paper-based case files is a documented practical challenge (Derechos Digitales, Dejusticia)
Commonly funded by
National / regional programmes
Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
Kela automates the pre-processing of documents submitted with benefit applications — classifying, converting and routing attachments — saving an estimated …