Semantic drift is one of the most expensive problems in enterprise AI today — and most organizations don’t know they have it. It’s what happens when “revenue” means something slightly different in your CRM than your data warehouse than your BI tool than your AI agent’s system prompt: as the definition of “revenue” drifts, your business’ ability to measure it accurately evaporates. This is exacerbated in AI systems: as the semantic layer drifts, hallucinations increase and token – compute – costs spiral. What AI needs is a structured, consistently defined semantic layer on top of your data.
The Apache Ossie (incubating) Apache 2.0, vendor-neutral YAML specification built in the open, with 50+ organizations collaborating to define metrics, dimensions, and AI context once, portably, across any tool. In this talk, we’ll go through the creation of the Ossie spec, how to get involved, and go through real-world examples of taming data for AI using Ossie.
