Beyond the Bus Factor: Managing Tribal Knowledge (brihatijain.com)

A new framework has been introduced to address "tribal knowledge" in engineering teams, emphasizing the need to preserve the intent behind decisions as systems evolve. This issue arises when essential knowledge is held by a few experienced engineers, leading to operational instability when these individuals leave or are unavailable. The framework proposes a combination of decision records, living documentation, and centralized repositories, supported by AI-driven LLM agents, to ensure critical knowledge remains accessible despite ongoing system changes. This approach is particularly significant for the AI/ML community as it seeks to mitigate "documentation entropy," whereby written knowledge becomes outdated and unreliable due to rapid code changes. By integrating lightweight mechanisms to capture and structure knowledge, the framework aims to reduce the dependency on individual memory and enhance organizational coherence. Key technical elements include Engineering Decision Records (EDRs) that document the reasoning behind design choices, living documentation generated from real-time code updates, and centralized repositories to streamline knowledge sharing and accessibility. Overall, this framework not only provides a path to better documentation practices but also leverages AI to maintain alignment and context in a fast-paced development environment.