Informatica's engineering team built CLAIRE, a multi-agent AI system embedded across the Intelligent Data Management Cloud (IDMC), to automate enterprise data workflows spanning discovery, governance, data quality, and pipeline orchestration. Single-agent approaches failed due to context overload, tool selection errors, and inconsistent outputs across multi-step workflows involving 50–60 model calls per request. The solution uses an orchestration agent as a control plane that routes tasks to specialized agents, each with focused toolsets and dedicated validation frameworks. Key engineering challenges included failure propagation across chained agents, dynamic dependency recomputation when users modify plans mid-execution, and building a semantic layer for intent decomposition. The data quality agent achieved a 90% task success rate, 98% grounding accuracy, and 1% hallucination rate. Workflows that previously took up to three months were reduced to days.
Sort: