---
type: concept
title: Layer 4 — AI & Intelligence
aliases:
  - ai layer
  - intelligence layer
  - ai providers
  - layer 4
modified: 2026-07-08
tags:
  - architecture
  - architecture-layer
  - ai
  - rag
  - suitecentral
---

# Layer 4 — AI & Intelligence

> Provider selection, AI configuration, task-aware providers, multi-agent orchestration, and the RAG/help knowledge base.

The multi-provider AI stack of the [[pages/concepts/suitecentral-code-architecture|seven-layer code architecture]]. AI decisions are made here — provider selection across multiple base providers plus local-model support, task-aware routing (field mapping, data quality, transformation suggestions), multi-agent orchestration, and the retrieval-augmented help knowledge base that powers the in-app help assistant. Confidence scoring and hallucination detection run in this layer before any output reaches [[pages/architecture/core-application-services|Layer 3]] — and every output is validated by [[pages/architecture/governance-safety|Layer 5]] before it is acted on.

## Questions this layer answers

- Which AI providers are supported?
- How does the help knowledge base answer questions?

Ask them live on the **[Code Architecture Dashboard](https://demo.kstratmdconsulting.com/code-architecture-dashboard.html)** — click the fourth layer of the infographic.

## Documentation & evidence

- [AI systems comparison](https://demo.kstratmdconsulting.com/docs/architecture/AI-SYSTEMS-COMPARISON.md) — the dual AI architecture and how the governed path differs from the legacy path
- [Help chat system](https://demo.kstratmdconsulting.com/docs/features/help-chat-system.md) — the RAG pipeline behind the in-app assistant
- Proof card: [AI providers](https://demo.kstratmdconsulting.com/docs/review/proof-cards/ai-providers.md)

## Related layers

- [[pages/architecture/http-api-edge|Layer 2 — HTTP-API Edge]] — all production AI traffic enters through governed proxy routes
- [[pages/architecture/governance-safety|Layer 5 — Governance & Safety]] — validates every AI output before it takes effect
- [[pages/architecture/core-application-services|Layer 3 — Core Application Services]] — executes what validated AI proposes

---

*Last refreshed: 2026-07-06*