Leading AI down the yellow brick road

Bringing AI into an engineering organization is a lot like hiring a junior developer. It can look things up, write some code, and even deploy a feature to production. Engineers are given SOPs, documentation, and golden paths so they do not stray too far off course. AI can act autonomously, but what happens if, as part of a deployment, an agent decides it should delete all data before making a schema change to a database? AI should be like any member of the engineering team. It should have access to SOPs, documentation, and, most importantly, golden paths to follow. AI also needs a golden path.