AI Risk, Security, and Regulation
A supplier’s email told the assistant to change the bank account, and it did. The approval screen showed twelve matched invoices, a total, and a button.
-
-
Evaluating AI Systems Before and After Release
A fictional claims assistant earns a higher judge score but misses required facts. Evaluation needs independent evidence, validated ratings, and outcome checks.
-
Shared AI Platform for Language Models
A fictional company shares model quotas across teams. Its failures show why capacity, versions, fallback, data handling, and costs need platform ownership.
-
AI Agent Tool Execution and Reliability
A fictional support agent repeats a refund after a timeout. The case shows why model-driven actions need stable operation IDs and enforced execution rules.
-
RAG Architecture and Answer Quality
A fictional HR assistant cites an outdated policy. Tracing its evidence reveals how ingestion, retrieval, permissions, and generation can fail together.
-
LLM Capabilities, Limits, and System Design
A fictional invoice pilot scores 97%, yet production exposes missing evidence, invented dates, and weak cost estimates despite schema-valid JSON.