AI-Native Organisations Run on Skills: How to Structure and Scale Them — Imad Touil, QuantumBlack
AI-native organizations need governed, shared skills registries to scale agentic workflows
QuantumBlack distinguished engineer Imad Touil outlines a two-loop agentic software stack — an inner loop (code agents with context manager, tools/MCPs, memory, skills loader) and an outer loop (workflows with skills, sub-agents, MCP servers, hooks) — supported by shared infrastructure like MCP gateways, model gateways, knowledge graphs, and workflow marketplaces. The core argument is that most teams create skills individually but very few govern and share them across the organization, which is the missing layer for true AI-native scale. This is a practitioner-facing architectural blueprint for enterprises trying to move from ad-hoc AI tooling to structured, governed agentic pipelines.