AB-100 decision sheets

• MCP onboarding wizard vs custom connector vs computer use vs dynamic ERP MCP server
• Data policies vs Agent 365 vs credit caps vs Azure Policy and Purview vs guardrail egress
• Model deployment guardrail vs agent guardrail vs user prompt attack vs document attack
• Prebuilt vs declarative vs custom engine vs Foundry Agent Service vs Azure infrastructure
• Classic vs generative orchestration in Copilot Studio
• Which data policy connector controls which Copilot Studio capability
• Pre-production evaluation vs red teaming vs continuous vs scheduled evaluation vs traces
• Theme vs knowledge-source vs topic vs quick vs test-chat test sets
• Retrieval vs few-shot prompting vs fine-tuning vs a custom model
• Azure OpenAI vs partner models vs dependent services vs monitoring in a Foundry cost case
• Prompt agent vs hosted agent vs Ephemeral agent (direct Responses API call) on Foundry
• Grounding vs not-found fallback vs inline citations vs system message vs temperature
• Analytics Viewer vs Bot Transcript Viewer vs Cost Management Reader vs Log Analytics Reader
• SharePoint connector vs file upload vs per-site registration vs endpoint filtering