Use ChatGPT Work to ground analysis in your semantic layer

# Use ChatGPT Work to Ground Analysis in Your Semantic Layer **TL;DR:** To trust ChatGPT for accurate data analysis, ground it in a shared semantic layer using data plugins, which can be built from scratch or integrated with existing systems like Snowflake Cortex or Databricks Genie, requiring sustained organizational effort. ## The Problem: Trust and Complex Data Logic A core question arises: how can we trust ChatGPT to correctly find and query data? Even seemingly simple business questions can involve very complex query logic or metric definitions. This critical knowledge is often scattered across different tools and team members. If an AI agent cannot access this information, it cannot make correct decisions. Therefore, the key is to enable the agent to access this information. Data plugins allow us to convert existing knowledge into shareable capabilities. ## Building the Semantic Layer as a Plugin The solution is to construct a **semantic layer** that encapsulates essential knowledge. This layer includes: * Key entities * Core metrics * Standard filtering dimensions * Open-ended questions and common pitfalls Team members familiar with the data must review and refine this layer. Once its quality is confirmed, it can be packaged as a plugin and shared across the entire organization. This approach is particularly valuable if you have a relatively clear business domain. ## Integrating with Existing Semantic Layers You are not required to build everything from scratch. If your team has already constructed a semantic layer—whether as an independent capability or within platforms like **Snowflake Cortex** or **Databricks Genie**—you can directly integrate it. The agent can learn from this existing knowledge base, eliminating the need to recreate content. ## The Organizational Imperative To truly succeed with this approach, it requires dedicated focus and the formation of a dedicated team to continuously optimize and build out the system. This work must be prioritized at an organizational strategy level. Its importance is understood, but it demands sustained attention and commitment. Further product developments in this area are forthcoming. Source: [OpenAI - Use ChatGPT Work to ground analysis in your semantic layer](https://www.youtube.com/watch?v=bbNTY8O0GTY)