Integrate agents with enterprise systems; Glimpse (NBA platform) /Veeva or Salesforce, Adobe and internal commercial platforms using Model Context Protocol (MCP), agent-to-agent (A2A) patterns, tool discovery, scoped authentication, and structured outputs. Build and maintain evaluation harnesses per agent, including golden datasets, calibrated LLM-as-judge scoring, and regression testing, and track agent KPIs covering task completion, grounding, reproducibility, human-override rate, latency, and cost per task.