Generative AI, large language models and agentic AI are the newest layers of the AI family. This lesson explains what each one is, how they fit inside machine learning and deep learning, and what makes a system "agentic" rather than just generative.
What generative AI creates
Generative AI (GenAI) is AI that creates new content: text, images, audio, video or code. Earlier ML models mostly predicted or classified. They answered "is this email spam?" or "what will sales be next month?". A generative model instead produces something that did not exist before, such as a draft email, a product photo or a summary of a long report. It does this by learning the patterns in huge amounts of training data and then producing new output that follows those patterns.
Foundation models and large language models
Most generative AI runs on foundation models (FMs). These are very large deep learning models, pre-trained on broad data, that can be adapted to many tasks instead of being built for one. A large language model (LLM) is a foundation model trained mainly on text, so it can understand and generate language: answering questions, summarising, translating or writing code. Most LLMs use the transformer architecture, a type of neural network that is good at relating each word to the words around it. So, put plainly, an LLM is a kind of foundation model, and foundation models are deep learning models.
From answering to acting
A chatbot built on an LLM answers when you ask. Agentic AI goes further: it is given a goal and works toward it with less step-by-step direction from a person. An agentic system typically:
- plans, breaking the goal into steps;
- uses tools, such as calling an API, querying a database or sending an email;
- remembers what it has done so far;
- adjusts when a step fails or new information arrives.
Most agentic systems today use an LLM as their reasoning engine, then wrap it with tools, memory and a loop that decides what to do next. Generating text is the agent's way of thinking; the tools are how it acts.
How the pieces nest
| Layer | What it adds |
|---|---|
| AI | The goal: machines doing tasks that need human-like intelligence |
| Machine learning | Learning those tasks from data |
| Deep learning | Many-layered neural networks, suited to unstructured data |
| Generative AI (FMs, LLMs) | Large pre-trained models that create new content |
| Agentic AI | GenAI models plus planning, tools and memory to act on goals |
Each layer builds on the one above it in the table.
On the exam
Watch for scenarios where a business wants a system to do something, such as update a booking, file a ticket or move money, not just say something. That is the signal for agentic AI. A bigger or better-tuned model alone does not let a system take actions; tools and orchestration do. Also remember the direction of the nesting: all generative AI is machine learning, but most machine learning is not generative.