Posts
All the articles I've posted.
AI That Cannot Be Tested Cannot Be Trusted: Building an Enterprise Evaluation System
Published: at 03:05 AMEnterprise AI needs more than demos and user anecdotes. Build a living evaluation system with task-specific tests, production signals, human assurance and release gates before confidence becomes operational risk.
Choose Your First Agentic Workflow by Reversibility, Not by Excitement
Published: at 02:20 AMA first production agent should be chosen for its ability to be observed, stopped and reversed. CIOs can learn faster from bounded workflows than from spectacular autonomous demonstrations.
The AI-Native Delivery System: Why Coding Agents Expose Weak Engineering Practices
Published: at 01:35 AMCoding agents do not fix weak engineering systems; they expose them. Leaders need smaller batches, usable context, safety nets and product feedback before increasing autonomous code delivery.