SCOI Atlas · L5 Execution
L5C Retrieval-Augmented Workflows
RAG in production — not the demo — including chunking, graph, freshness and provenance.
TL;DR · Direct answer
L5C Retrieval-Augmented Workflows is a sublayer of L5 Execution in the Supply Chain of Intelligence™ (SCOI) by Anand Arivukkarasu. It is a required capability but not on its own a durable moat. 'RAG works in demo, breaks in production' is almost always a freshness or permissioning problem, not a model problem.
What actually matters at L5C
- Freshness policy is where enterprise RAG deployments live or die.
- Graph RAG > vanilla RAG for anything with entity structure.
- Access-controlled retrieval is a compliance requirement, not a feature.
The startup lens
'RAG works in demo, breaks in production' is almost always a freshness or permissioning problem, not a model problem.
Vertical lens — how this plays across categories
Enterprise search
Permissions-aware retrieval + graph enrichment.
Legal / research
Provenance-cited output as the required format.
Support AI
Freshness-first retrieval; stale answers destroy trust.
How to defend L5C
- Own the retrieval graph, not just the vector index.
- Cited output as a product default.
Other sublayers in L5 Execution
L5A ★
Domain Execution & Tool Use
The domain-specific tools an agent can actually call — and the semantics of each tool.
L5B ★
Decision Frameworks & Reasoning Scaffolds
The prompted or code-enforced structure that turns raw model reasoning into a decision.
L5D ★
Operating Playbooks
Codified 'how work gets done here' — for agents to execute, humans to review, and both to improve.
L5E
Interaction Skills & Actuation
The 'hands' of the agent — browser use, keyboard-mouse, robotic actuation, physical control.