Architect agentic systems, RAG pipelines, and AI-augmented workflows across the Microsoft, OpenAI, and Anthropic ecosystems (Azure OpenAI, Azure AI Foundry, AI Search, Copilot Studio, the OpenAI platform including Responses API and Agents SDK, and the Anthropic platform including Claude API, Claude Agent SDK, MCP, and skills), choosing and composing them based on client fit, not vendor loyalty. You have built agentic systems with real tool use (MCP servers, skills, sub-agents, multi-step planning and reflection loops, or comparable patterns on OpenAI Agents SDK, Claude Agent SDK, Semantic Kernel, or similar) and can walk us through the architecture and the failure modes you hit.