Abstract:
As legal practitioners in Indonesia conduct research and draft documents they
face the daunting task of sorting out the vast array of existing laws. Coming up against
the limits of a manual approach, the process of cross-referencing and drafting compliant
legal documents can be extremely time consuming. The lack of tailored insights and
localised information from existing tools exacerbates the problems, and for this reason,
the “Legal Researcher AI Assistant” initiative has suggested an AI-powered chatbot that
makes use of a Large Language Model. The goal of this system is to bring a real boost
in speed and accuracy to the legal research process, and also takes on the jobs of drafting
legal documents and dissecting legal problems. Well-known for its scalability and
effectiveness, the chosen answer is a Multi-Agent System, which takes a modular
approach with specialist AI agents for different tasks, for instance, research, drafting
and opinion generation. Its Retrieval-Augmented Generation (RAG) setup stores
summaries of laws in a vector database and pulls the full texts from government sources
as and when needed, so that the system stays accurate. The coordination by LangGraph
and the modular design of the system makes it cost-efficient too. The project follows a
continuous cycle of requirements gathering, design, implementation and rigorous
testing, and will ensure that its product is reliable and practical.
