AI Data Supply Chains Should Start With Your Systems of Record, Argues Craig Mullins

craigmullins · x · 2026-09-18

In a TechChannel article, Craig Mullins argues enterprise AI discussions over-index on downstream components (LLMs, vector databases, RAG, agents) while neglecting the upstream question of data provenance. Borrowing the manufacturing supply-chain metaphor, he contends organizations need an end-to-end AI data supply chain that traces information from origin to application, ensuring it is trusted, governed, and delivered with context — and that this chain should begin with the enterprise system of record, not with model selection.

Original post →

More from Infra

Infra channel →