Transforming Legal Analysis: MANZ Integrates AI Research Tools

MANZ recognized the potential of generative AI and large language models (LLMs) to revolutionize legal research, addressing challenges like outdated methods, specialized legal terminology, and iterative discussions. With AWS Partner deepset collaborating with Amazon Web Services (AWS) and AWS Partner Anthropic, MANZ developed an AI-powered legal assistant designed for transparency and efficiency, enhancing legal workflows, and ensuring informed decisions. Beta-tested by 200+ experts in 2024, it delivered 20% higher accuracy than traditional tools.

Lawyers, judges, and other legal professionals must build a well-rounded picture of a case by looking at related incidents, making sure that no precedent escapes attention. Depending on the complexity of the case, this can require reviewing hundreds or even thousands of documents— a considerable investment in time and resources. MANZ recognized the transformative potential of LLMs for the legal sector but faced uncertainty in how to apply this technology effectively. Their vision was to make legal research more efficient and accurate through an AI-powered assistant capable of conducting complex searches, answering intricate legal questions, generating summaries, and drafting contracts.

However, the intricate nature of legal documents, characterized by specialized terminology and nuanced interpretations, posed significant obstacles. Developing a solution required addressing the complexity of legal nomenclature, maintaining transparency in data sourcing, and ensuring the ability to manage iterative discussions effectively. Keeping information current and reliable added another layer of difficulty to the project. These challenges underscored the need for a delicate balance between cutting-edge AI capabilities with a nuanced understanding of legal contexts. This required close collaboration between MANZ’s development team and deepset’s team of AI experts to create an industry-customized tool that could truly transform legal research practices.

To address their challenges, MANZ collaborated with deepset to develop MANZ Genjus KI, an advanced AI-powered legal assistant built on the deepset AI platform—a comprehensive LLM-application and agent development platform that leverages AWS infrastructure with a seamless integration with Amazon Bedrock. This enables the platform to use Amazon Bedrock as an LLM inference backend, allowing customers to use models cost-efficiently at scale. The project began with building a comprehensive knowledge base. MANZ’s publishing content was stored using deepset in Parquet files on Amazon Simple Storage Service (Amazon S3), forming the foundation of the AI system. The deepset platform uses an event-driven system to process data added to Amazon S3, processes it in a scalable data indexing process in a Kubernetes cluster and stores data alongside vectors in Amazon OpenSearch Service to make it available to the retrieval-augmented generation (RAG) system. Semantic data, such as text similarity embeddings, is integrated natively into the deepset platform using Amazon OpenSearch Service, which serves as both a vector store and a search engine to enhance document retrieval performance. This setup enabled precise semantic search and retrieval, ensuring the AI assistant could process and respond to complex legal queries.

The deepset platform was also used as the primary prototype and development environment running on AWS, allowing MANZ and the deepset team to explore use cases, build with models, and iterate efficiently. The infrastructure was further supported by deepset’s integration with Amazon Redshift, used as a data warehouse for structured data. deepset’s integration with Amazon Bedrock played a pivotal role in enabling MANZ to harness the power of Visual Language Models (VLMs) and LLMs for processing unstructured data and handling complex legal queries. The system streamlined data processing, using optical character recognition (OCR) for text conversion, VLMs for image and table captioning, and automated metadata generation. Using deepset as the AI orchestration layer, MANZ securely accessed Anthropic’s Claude 3.5 Sonnet via Amazon Bedrock, using it to intelligently generate responses to user queries based on MANZ’s data. With deepset’s Pipeline Builder, they created a sophisticated AI pipeline to further enhance inputs and outputs of the LLM, delivering excellence in re-ranking, extraction, summarization, and source verification. The pipeline formed the foundation of a RAG chatbot that uses embedding and vector database management to address intricate legal inquiries, support iterative discussions, and maintain contextual understanding across sessions. deepset further enhanced this infrastructure with tools for prompt engineering, ensuring precise and consistent responses. By combining seamless LLM integration with scalable, flexible infrastructure, MANZ Genjus KI now empowers legal professionals with innovative tools for smarter, faster decision-making — offering advanced features such as AI-powered searches, legal decision summarization, and contract drafting while maintaining data integrity and transparency.

MANZ Genjus KI has revolutionized legal research with AI-powered capabilities, including advanced search, complex query handling, and legal decision summarization. Beta-tested by over 200 experts using deepset’s turnkey testing capabilities in Q2 2024, the product demonstrated a 20% improvement in accuracy compared to existing tools. The solution efficiently retrieves documents while addressing complex legal queries, all while maintaining state across iterative discussions.

The product addresses key legal challenges with tailored AI capabilities. deepset’s platform provided the tools to ensure transparency into full source documents, providing references critical for legal professionals. The embedding model component in the deepset pipeline is fine-tuned for legal terminology, enabling precise processing of specialized language. Continuous database updates ensure users access current case law and avoid outdated information. Designed for iterative use, the deepset platform tracks query history, supporting complex research workflows. Building on its powerful foundation, MANZ Genjus KI improves legal workflows by automating tasks like legal drafting, such as contract creation, harnessing both specialized and general AI.

MANZ is transforming the legal field with a customer-centric approach that emphasizes accuracy, efficiency, and constant iteration. By leveraging deepset’s platform to combine a specialized legal knowledge base with cutting-edge AI and industry-specific customizations, the solution empowers legal professionals to conduct faster, more accurate research. With its innovative features and commitment to ongoing refinement, the product is poised to become an indispensable tool in the legal community, enabling informed decisions while streamlining workflows.

Founded in 1849, MANZ is the Austrian market leader for legal information. It offers a wide range of literature for professionals in the field of law and taxes, in print as well as digital formats. MANZ offers a variety of digital solutions and services, including secure electronic remittance of legal documents. MANZ customers are lawyers, law courts, notaries, authorities, lobbyists, business and real estate trustees, companies of all sizes, students of the law and economics, as well as scholars.