agent design

An opinionated look at common enterprise mistakes with agentic AI, from unbounded autonomy to weak observability and unclear ownership.

A countdown of common AI agent security mistakes, from shared keys and overprivileged identities to prompt injection, weak egress control, and missing audit trails.

A practical guide to replacing API keys with managed identities for AI agents that call Azure OpenAI, Azure AI Search, Key Vault, Storage, and other tools.