Screaming in the Cloud Fixing Shadow AI and Surviving re:Invent with Chase Douglas
Feb 5, 2026
Chase Douglas, CEO of Archodex and former AWS principal/CTO-level engineer, builds AI governance and compliance tools for enterprises. He talks about Shadow AI risks and why centralized pipelines matter. He describes running isolated accounts for experiments, chaotic model availability and pricing, and why re:Invent felt stranger this year. He ends upbeat about spec-driven AI coding workflows.
AI Snips
Chapters
Transcript
Episode notes
Centralize AI Interactions Through A Pipeline
- Centralize all AI interactions through a governed pipeline to handle audit, anonymization, and analytics.
- Capture who accesses logs and route data to the right destinations before scaling AI features into products.
Superfund: An Isolated Playpen For AI Experiments
- Corey runs Claude Code in an isolated AWS account named "Superfund" with admin role but no persistent sensitive data.
- Everything must exit via a guarded CICD pipeline to prevent accidental leakage into core environments.
Shadow AI Creates Hidden Compliance Risk
- Shadow AI arises when developers use external models and services outside corporate guardrails during spikes and POCs.
- This creates unknown logging, data mixing, and compliance risk that central pipelines must detect and capture.