Initiative overview
The Brazilian Chamber of Deputies provides a wide range of internal services that support roughly 10,000 people, including 513 deputies, parliamentary staff, public servants and outsourced workers. Its operations span broadcasting (radio and TV), news services, medical care, engineering, architecture, finance, procurement, IT, legal services, training, a legislative university, dozens of legislative committees and extensive research and library functions. This diversity generates a constant demand for rapid access to accurate institutional information.
The Chamber’s traditional intranet model could no longer meet the speed and analytical depth required by employees in their daily work. As generative AI became widely used, staff increasingly turned to external platforms; however, using these tools without institutional context would require uploading large volumes of internal documents, creating risks around data privacy, rule misapplication and high duplication‑related costs. To address this, the Chamber developed Ulysses Chat—a generative AI‑based virtual assistant designed specifically for internal use and grounded in the institution’s own knowledge base.
Ulysses Chat draws on a network of interconnected AI agents and operates within a governance layer designed to ensure privacy, security and responsible AI behaviour. It incorporates mechanisms for auditing, token accountability, minimising bias and hallucinations, and managing out‑of‑context queries. In its first implementation stage, the system answers questions using official institutional data, including internal standards, legislation, manuals, administrative rules, planning and budget information, logistics and human‑resources procedures, and legislative data such as bills, speeches and voting records. It also uses the Chamber’s organizational chart to contextualize information. In the current second phase, Ulysses Chat is beginning to automate selected internal processes through direct interaction with institutional systems. Across all information domains, data managers, data stewards and the Data Protection Officer oversee ongoing data‑quality monitoring.
