Field Note — AI Systems Architecture

SYSTEMS NOTE // 04  ·  JULY 2026  ·  SGI RESEARCH

WHAT IS AN
AGENTIC AI?

Chatbots, RPA, and RAG all look autonomous from the outside. None of them are. This note separates the three patterns routinely mistaken for agentic AI from the one architecture that actually qualifies — a loop that discovers, plans, delegates, and revises.

Not agentic Agentic

Click a legend tag to spotlight that block

PINNED0Patterns mistaken for agentic AI
PINNED0Core capabilities inside the loop
PINNED0Specialist agents coordinated
PINNED0Feedback loop closing the system
LLM Chatbots◆ RPA◆ RAG◆ Memory◆ Tools◆ Planning◆ Feedback◆ Orchestrator LLM◆ Multi-Agent Protocol◆ LLM Chatbots◆ RPA◆ RAG◆ Memory◆ Tools◆ Planning◆ Feedback◆ Orchestrator LLM◆ Multi-Agent Protocol◆

Block A — Non-Agentic Patterns

Three Patterns Mistaken
For Agentic AI

LLM Chatbots
Query
System Prompt
LLMs
Output
RPA
Query
Script Trigger
Only Execute Tools
LLMs (Use Tools)
Output
RAG
Query
Embedding
Knowledge Base
Vector Db
Web Search Data
Data Augmentation
LLMs
Output

A SCRIPT THAT CALLS TOOLS
IS NOT AN AGENT.
A MODEL THAT RETRIEVES
CONTEXT IS NOT AN AGENT.
NEITHER DECIDES WHAT
TO DO NEXT.

Block B — Agentic AI System

Inside An
Agentic Loop

PINNED
AGENT_01 Core Loop Orchestrator + 4 Capabilities
Orchestrator LLMDirects the loop. Reads the query, decides which capability to invoke, and arbitrates between memory, tools, planning, and feedback
MemoryRetains prior turns and intermediate results so decisions build on what already happened
ToolsExecutes external functions — code, APIs, search — chosen by the orchestrator rather than hard-coded in advance
PlanningBreaks the query into an ordered sequence of sub-tasks before any tool is called
FeedbackEvaluates the outcome of each step and revises the plan — the mechanism a static pipeline lacks entirely
PINNED
PROTOCOL Multi-Agent Protocol Coordination Layer
DiscoverIdentifies which specialist agents exist and what each one is capable of doing
Share TasksRoutes sub-tasks from the orchestrator to whichever specialist agent is best suited to handle them
Update Task InfoKeeps task state synchronized across agents as work moves back and forth through the loop
PINNED SPECIALIST_01 Coding Agent

Writes, edits, and executes code sub-tasks handed off by the orchestrator.

PINNED SPECIALIST_02 Retrieval Agent

Pulls grounding data from knowledge bases, vector stores, or the web on request.

PINNED SPECIALIST_03 Citation Agent

Traces claims in the output back to their sources before the answer is returned.

↻Output feeds back into Query — the loop that a linear pipeline never closes

IF IT CAN'T LOOP,
DISCOVER, AND REVISE —
IT'S A PIPELINE,
NOT AN AGENT.

More Systems.
More Notes.
Legend Red — not agentic Green — agentic