monday.com has developed an AI Slack bot to streamline how developers access platform data. The initiative addresses a significant gap in data accessibility for developers, who previously had to manually check various internal tools for information on service statuses or blocked accounts. The bot allows users to query specific service statuses directly in Slack, which is the primary interface used by monday.com developers. The architecture behind the bot involves several components. It uses a new service to connect with Slack and an LLM model that processes user queries. The Model Context Protocol (MCP) enhances communication with the AI model, functioning like a USB port for data. The bot processes messages through a series of steps, enriching them with context before querying the model for real-time information. Developers opted for open-source frameworks like LangChain to build the bot quickly, allowing for flexibility and rapid prototyping. Challenges arose due to the non-deterministic nature of AI. The bot's outputs can vary significantly with the same input, influenced by parameters like temperature and top-p. To mitigate inaccuracies, the team decided to minimize creativity by setting the temperature to zero for their specific analytics workload. They also learned not to rely on the LLM for calculations, pre-computing necessary data on the server instead. The team has since found additional use cases, such as automating account block suppressions while ensuring human oversight through a Human-in-the-loop pattern.