The 2026 Enterprise AI Stack That Actually Ships
After deploying AI at 6 enterprises, the boring stack that compounds.
Long-form pieces on the operational decisions that actually matter in 2026.
After deploying AI at 6 enterprises, the boring stack that compounds.
After running RAG for a Fortune 500 helpdesk, here's what actually works.
Real AI governance starts with eval pipelines, not policy documents.
After negotiating both, the unit economics and feature differences.
Bad chunking. Bad eval. Bad retrieval. The patterns that survive prod.
The data platform decision when AI is on the roadmap.
Langfuse + MLflow + custom evals. The 2026 production stack.
What actually ships. What's still demo-ware. The honest 2026 take.
When self-hosted Llama 3 wins, when GPT-5 API wins. The math.
Why boring beats clever for enterprise AI deployments.