evidoria

← Back to browse

Good practice

Amauta.Pro — Peru's AI Assistant for Drafting Domestic-Violence Protection Orders

Peru · Lima · See the Peru profile

Peru's Lima Norte court built Amauta.Pro, an AI assistant that drafts domestic-violence protection-order resolutions; the court reports draft time fell from over three hours to 40 seconds, with 1,000+ resolutions produced in its first weeks.

>3 hours
Time to draft a resolution, before Amauta.Pro (pre-September 2024)
~40 seconds
Time to draft a resolution, with Amauta.Pro (since September 2024)
1,000+ resolutions
Resolutions produced (first weeks of use)
23,000 measures
Protection measures issued, Lima Norte jurisdiction (2022)
~80 %
Share of 2022 protection measures classified mild/moderate risk (2022)
545 activations
Panic Button app activations (Jan-Jun 2024)
199 devices
Panic Button devices in use (Jan-Jun 2024)
Amauta.Pro — Peru's AI Assistant for Drafting Domestic-Violence Protection Orders

Details

Promoter
Corte Superior de Justicia de Lima Norte (Lima North Superior Court of Justice), Poder Judicial del Perú, developed with the Universidad Nacional de Ingeniería (UNI)
Period
Severity-triage model from 2022; Amauta.Pro presented September 2024; operational in Lima Norte family-violence courts since 2024
Keywords
judiciary, gender-based violence response, natural language generation, e-justice

Context

The Superior Court of Justice of Lima Norte, part of Peru's Poder Judicial, handles a heavy domestic-violence caseload — in 2022 its jurisdiction issued 23,000 protection measures, about 80% classified as mild or moderate risk under a severity-triage model the court built with the Universidad Nacional de Ingeniería (UNI). Building on that model, judges and engineers at the court's in-house AI laboratory spent about a year developing Amauta.Pro, an AI assistant that reads a domestic-violence complaint and the documents submitted through family police stations, organises the case file, applies the applicable legal norms, and drafts a protection-order resolution.

Objectives

The goal, as described by court president Lorenzo Castope Cerquín, was to automate classification and drafting for the large share of simple, lower-severity cases so judges could concentrate on severe and very-severe ones, targeting a four-hour turnaround for issuing protection measures.

Activities

Amauta.Pro was presented on 5–6 September 2024 at the II International Congress on Artificial Intelligence in the Justice System, under Judicial Power president Javier Arévalo. Drafts produced by the tool are reviewed and signed by the presiding judge before issuance — the system assists rather than replaces judicial decision-making.

Results

According to the Poder Judicial's own figures, reported independently by Infobae and other Peruvian outlets, generating a draft resolution now takes about 40 seconds versus more than three hours previously, and the tool produced more than 1,000 resolutions in its first weeks of use. The same court's separate 'Panic Button' app for domestic-violence victims logged 545 activations across 199 devices between January and June 2024.

Conclusions

No independent audit of Amauta.Pro's resolution quality, error rate or bias has been published — all figures to date originate from the Judicial Power's own announcements as covered by Peruvian press, not a peer-reviewed evaluation. Court officials have described the tool as replicable to other case types, but no documented expansion beyond Lima Norte's family-violence courts had been reported as of mid-2026.

Implementation

Indicative cost
Low (< €50k) — No budget figures are published. The tool was developed in-house by court judges and engineers over about one year, suggesting a low-cost, internally resourced initiative rather than a large procurement.
Time to results
Short (< 1 year) — About one year of development; presented September 2024; in production use at Lima Norte as of mid-2026 with no reported expansion.
Staffing & skills
In-house AI laboratory at the Superior Court of Justice of Lima Norte — judges and engineers, Universidad Nacional de Ingeniería (UNI) — partner on the earlier severity-triage model, Court president Lorenzo Castope Cerquín — sponsor, Presented under Judicial Power president Javier Arévalo

Conditions for success

  • Human-in-the-loop design: drafts reviewed and signed by the presiding judge before issuance
  • Built on an existing severity-triage classification model rather than starting from scratch
  • Strong institutional sponsorship within the court

Common failure modes

  • No independent audit of resolution quality, error rate or bias has been published
  • All performance figures are self-reported by the Judicial Power via press coverage, not peer-reviewed evaluation
  • No documented expansion beyond Lima Norte as of mid-2026 despite claims of replicability

Where it fits

Governance type
judicial branch, single-court in-house development
Scale
subnational — one court's jurisdiction
Income level
upper-middle-income (Peru)

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful