Real-World Agentic Systems (Under the Hood)

So far, we’ve covered all the ingredients that make up an agent:
tools, planning, RAG, memory, structure, and coordination in multi-agent setups.

But you might be thinking:

“Where does all this actually show up in the real world?”

Let’s walk through a few public-facing systems that exhibit agentic behavior — as far as we can tell.

⚠️ Note:
These aren’t open source. We don’t know their exact internals.
What follows is an informed simplification based on how they behave externally — just enough to understand how the agentic stack might show up in practice.

NotebookLM (Google): Agentic Search on Your Own Data

Google’s NotebookLM acts like a personal research assistant. You upload your files, and it helps you work with them — summarizing, answering questions, even generating audio versions or study guides.

Core focus: Q&A over your content — essentially a scaled-up, personal RAG system.

How it likely works:

User uploads files (PDFs, notes, slides, etc.)
Preprocessing — Stores them for retrieval later.
User asks a question — e.g., “What were the key insights from my Q2 strategy deck?”
Planning — Interprets task type (summary, Q&A, comparison?), identifies relevant docs/sections.
RAG — Retrieves the most relevant document chunks.
LLM Generation — Responds clearly, grounded in your content.
Memory —
- Short-term: Tracks the conversation.
- Long-term: Likely minimal or none.
Tools — Possibly file viewers, summarization modules.

What makes it agentic: Interprets goals, searches across your data, and composes responses — not just static outputs.

Perplexity: Agentic Search on the Open Web

Perplexity gives you a direct, answer-like response with sources — instead of a page of links.

How it likely works:

User asks a question — e.g., “What’s the latest research on Alzheimer’s treatments?”
Planning — Interprets intent (“latest,” “credible”), decides search approach.
Tool Use — Issues queries via web APIs.
RAG — Retrieves relevant page snippets.
LLM Response — Synthesizes an answer with citations.
Memory —
- Short-term: Session context.
- Long-term: May store preferences (e.g., “always use WSJ for news”).

What makes it agentic: Fetches info, decides what to use, and constructs an answer in a multi-step loop.

DeepResearch (OpenAI): Deep Agentic Workflows

DeepResearch tackles open-ended, complex research tasks — e.g., market analysis, competitive landscapes, technical deep dives.

How it likely works:

User asks a broad task — e.g., “Analyze the generative AI landscape for education startups.”
Planning — Breaks into subtasks (funding, trends, companies, risks), forms an execution plan.
Tools — Likely includes:
- Web search
- Document readers (PDFs)
- Data tools (spreadsheets, graphs)
- Report generation modules
Agentic RAG — Not one-shot retrieval — fetches, reflects, re-fetches as task evolves.
Memory —
- Episodic: Tracks which parts are done.
- Semantic: Stores key facts/names.
Multi-step Reasoning — Loops: plan → retrieve → read → rethink → generate → refine → repeat.