evidoria

← Back to browse

Good practice

Ulysses — Brazil's Chamber of Deputies institutional AI programme for legislative data and staff support

Brazil · Brasília · See the Brazil profile

Since 2018 Brazil's Chamber of Deputies has run Ulysses, an AI programme that automates legislative-data workflows; a 2025 relaunch made it a formal institutional programme with new AI-governance rules, staff access to major LLMs, and a chatbot in development.

~300,000
Historical information requests used to train Ulysses's routing model (logged since 2003)
22
Subject-matter knowledge areas Ulysses routes requests across
14
Parliamentary chambers co-leading the IPU AI-governance collaborative
Ulysses — Brazil's Chamber of Deputies institutional AI programme for legislative data and staff support

Details

Maturity
Scaling
Promoter
Câmara dos Deputados — Diretoria de Inovação e Tecnologia da Informação (Ditec)
Period
2018–present
Keywords
legislative data, NLP, generative AI, parliamentary innovation, governance

Context

Brazil's Chamber of Deputies (Câmara dos Deputados) launched Ulysses in November 2018 — a set of AI algorithms, built by the Chamber's Directorate of Innovation and Information Technology (Ditec), that automatically articulates and organises legislative data across the Chamber's portal and internal systems. The tool is named after Ulysses Guimarães, president of Brazil's 1987-88 Constituent Assembly.

Objectives

Automate and organise legislative-data workflows, and from 2019 automatically distribute incoming parliamentary information requests across subject-matter knowledge areas to reduce the manual effort previously needed to triage this repetitive workload.

Activities

From 2019, the Chamber's Legislative Consulting office (Conle) put Ulysses to operational use, distributing incoming information requests among 22 subject-matter knowledge areas, trained on roughly 300,000 historical requests logged since 2003. On 11 December 2025, Chamber President Hugo Motta relaunched Ulysses as a formal institutional AI programme alongside newly published AI-Governance and Data-Governance policies; staff gained sanctioned access to Claude, Gemini and GPT, and the Chamber announced an internal assistant, 'Ulysses Chat', plus planned features to route bills to committees and flag potential constitutional conflicts. The Chamber also co-leads, with the European Parliament, a 14-chamber collaborative under the Inter-Parliamentary Union's Centre for Innovation in Parliament, which published shared AI-governance guidelines for parliaments worldwide in December 2024.

Results

Ulysses has been in continuous operational use since 2019, automatically distributing real parliamentary information requests; the Chamber reports this reduced the manual effort previously required for triage, though no independent quantitative evaluation of time or error savings was found.

Conclusions

Secretary-General Lucas Ribeiro stressed that human review applies at every stage of AI-assisted work. The programme has grown from a single 2018 tool into an institution-wide effort with both domestic policy weight and an international governance-leadership role.

Implementation

Indicative cost
Medium (€50k–€500k) — Not itemised in the source; involves in-house Ditec development plus licensed access to external commercial LLM platforms (Claude, Gemini, GPT) for staff.
Time to results
Long (> 3 years) — Initial tool launched November 2018; operational request-routing use from 2019; relaunched as a formal institutional programme with new governance policies in December 2025.
Staffing & skills
Ditec (Directorate of Innovation and Information Technology) engineers, Conle (Legislative Consulting office) staff using the request-routing tool, Secretary-General oversight of AI governance

Conditions for success

  • Newly published AI-Governance and Data-Governance policies
  • Mandated human review at every stage of AI-assisted work
  • Sanctioned, controlled staff access to external generative-AI platforms

Common failure modes

  • No independent quantitative evaluation of time or error savings published
  • Underlying model architecture not publicly disclosed

Where it fits

Governance type
national legislature (lower house)
Scale
national institution
Income level
upper-middle-income

Data sources

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

Attachments

Similar practices you may find useful