L23 · From Chatbot to AI System: The Model vs. the Product

L23 · Phase 6 · 25 min

From Chatbot to AI System: The Model vs. the Product

“When I chat with AI Tool, am I talking directly to 'the model', or to something built around it?”

AI System Architecture Builder

Back in Tools, you built a Kitchen Inventory Agent for yourself — a chat interface, a rules engine, and a layer that turns what you say into a structured update. It's a product built around an LLM, not the LLM itself. Assemble the correct order its own layers actually run in.

Your order will build here…

Prototype note: this is the Kitchen Inventory Agent’s own architecture, simplified to six labelled stages — real production systems often add more layers (logging, safety checks, retries) between these.

Depth ladder

The chat window you type into is a product built around a model, not the model itself — the same way a car's dashboard is built around an engine. The Kitchen Inventory Agent makes this concrete: its interface, its rules engine, and its intent schema all sit around the underlying LLM, not inside it.

Knowledge check

Given a feature list (memory, voice input, refusal behaviour, the underlying LLM), sort each into 'model' or 'product'.