# Bliki: Using LLMs to Generate Specialized Organizational Glossaries Improving LLM performance in specialized domains by creating structured, context-aware organizational terminology references. - **Source:** [Martin Fowler](https://martinfowler.com) - **Published:** 2026-11-12 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/64622004-fdbf-4c75-8166-dc0e057f3546 ## Addressing Specialized Vocabulary Gaps Large organizations rely on unique terminology that LLMs often fail to recognize because these terms are absent from general training sets. By using an LLM to scan internal documents, emails, and chat logs, companies can identify specialized vocabulary. This process ensures that AI agents operate with a deeper understanding of internal organizational language. ## Identifying Contextual Term Usage The glossary generation process must capture both new terms and existing words used in unconventional ways. LLMs should be instructed to provide alternative definitions when a term carries different meanings across departments. This approach clarifies ambiguity, ensuring that AI agents interpret specific jargon correctly based on the relevant operational context provided. ## Human Oversight and Validation Draft glossaries generated by AI require rigorous review by human domain experts to correct inaccuracies or misinterpretations. Once verified, these glossaries should be stored in version control systems to track changes over time. Regular refresh schedules involving both human and agentic input maintain the relevance and accuracy of the terminology database. ## Managing Bounded Contexts Organizations should avoid creating a single, monolithic glossary, as terminology often shifts between different operational areas. Instead, implementing the Domain Driven Design concept of bounded contexts allows for multiple, specialized glossaries. This structure enables AI agents to consult only the relevant definitions required for their specific tasks, improving overall system precision. ## Optimizing Token Consumption For large glossaries, efficiency is critical to minimize token usage during AI processing. Developers should include only essential terms and links to full definitions within the initial context window. Models can then selectively retrieve detailed explanations only when necessary, balancing the need for comprehensive knowledge with the constraints of model performance. ## Applications in Agentic Engineering Glossaries serve as foundational tools in agentic engineering, particularly for generating consistent code and documentation. By aligning software naming conventions with enterprise vocabulary, organizations ensure internal consistency. Sensors can be integrated into development harnesses to monitor compliance, ensuring that generated outputs adhere to the established terminology standards across all internal systems. ## Preventing Internal Term Exposure Effective glossary management includes identifying terms that should remain internal and not appear in customer-facing interfaces. By explicitly flagging deprecated or internal-only language, organizations prevent confusion for external users. This practice maintains a clear distinction between the internal bounded context and the language presented to the end customer during interactions. ## Strategic Value of Ontologies Constructing explicit context layers through domain modeling and ontologies addresses the common struggle of specialized language in large enterprises. While generating glossaries is a relatively simple task, it acts as a vital component of broader data strategy. This systematic approach significantly enhances the effectiveness of AI agents operating within complex organizational environments. ## Sources covering this story - https://martinfowler.com/bliki/GeneratedGlossary.html — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._