evidoria

← Back to browse

Good practice Imported

Argentina's National Pilot Cuts Judicial Task Times by Up to 90% with Generative AI

Argentina · Buenos Aires · See the Argentina profile · See the Buenos Aires profile

Evidence: Descriptive / self-reported Top 32% 67/100 · Ask Evidence Copilot about this practice

Argentina's Ministry of Justice and SAIJ piloted generative AI across the judiciary: 5,000 court staff rated it 81% positive, and identifying grounds for appeal fell from 25 minutes to 2.5, with jurisprudence searches cut by around 90%.

~5,000
Judicial staff surveyed (2024-2025)
81%
Positive rating of the experience (2024-2025)
25 minutes -> ~2.5 minutes
Time to identify grounds for appeal (agravios), before vs. AI-assisted (2024-2025)
~77%
Quality benchmark retained on AI-assisted task (2024-2025)
~90%
Reduction in jurisprudence search time (2024-2025)
1,007 (three trials each), ~3,000 prompts
Structured test cases run (2024-2025)
Argentina's National Pilot Cuts Judicial Task Times by Up to 90% with Generative AI

Details

Maturity
Pilot
Promoter
Ministerio de Justicia de la Nación / Sistema Argentino de Información Jurídica (SAIJ)
Period
2024-2025
Keywords
justice, judiciary, generative AI, public administration, legal research

Context

Argentina's judiciary sought to test whether generative AI could speed up common analytical tasks such as drafting and legal research across courts nationwide.

Objectives

Pilot the strategic and responsible use of generative AI with judges, prosecutors and court staff, measuring both user sentiment and task-completion time/quality on structured test cases.

Activities

Argentina's Ministerio de Justicia, through SAIJ and the National Program of Artificial Intelligence in Justice, ran the pilot in 2024-2025 with university AI-lab partners (UBA-IALAB, ALTA, JusLab, UNLZ Innovation Laboratory), running 1,007 structured test cases (three trials each) and roughly 3,000 prompts, and surveying close to 5,000 members of the judicial system.

Results

81% of surveyed judicial staff rated the experience positively. On identifying the grounds (agravios) in a recurso de amparo, AI-assisted work cut completion time from 25 minutes to about 2.5 minutes while retaining roughly 77% of the quality benchmark, and jurisprudence searches were sped up by around 90%.

Conclusions

The pilot's preliminary, self-reported figures are promising indicators of time savings and positive reception, but the source material does not disclose error or false-positive rates, so results should be read as pilot-stage indicators rather than an independently validated impact assessment.

Implementation

Indicative cost
Low (< €50k) — A software pilot using generative AI in partnership with university labs rather than bespoke infrastructure build-out, consistent with a low-cost pilot-stage deployment.
Time to results
Short (< 1 year) — Pilot ran through 2024-2025, with preliminary results presented in April 2025.
Staffing & skills
Ministerio de Justicia de la Nación / SAIJ programme team, university AI-lab partners: UBA-IALAB, ALTA, JusLab, UNLZ Innovation Laboratory, participating judges, prosecutors and court staff nationwide

Conditions for success

  • Partnership with multiple university AI labs to design and validate the pilot methodology
  • Structured testing protocol (1,007 test cases, three trials each) rather than anecdotal use
  • National-scale survey of nearly 5,000 judicial staff to gauge reception

Common failure modes

  • Results are self-reported by the implementing programme rather than independently audited
  • Error or false-positive rates for the AI outputs are not disclosed in the source material

Where it fits

Governance type
national justice ministry / judiciary
Scale
nationwide pilot, ~5,000 staff surveyed
Income level
upper-middle-income

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.

Attachments

Similar practices you may find useful