Skip to main content

Lesson 7: Generative vs. Agentic AI

Topics Covered
  • Generative AI: Reactive systems that create content.
  • Agentic AI: Proactive systems that pursue goals.
  • The Role of LLMs: Cognitive engines for both approaches.
  • Chain of Thought: How agents break down complex problems.

What is the difference between these two approaches?

1. Generative AI (The Creator)​

Generative AI systems (like chatbots and image generators) are fundamentally reactive.

  1. Wait: They wait for a user prompt.
  2. Generate: They produce content (text, code, images) based on learned patterns.
  3. Stop: Their work ends at generation. They do not take further steps without input.

Example: A content creator using AI to write a script. The human directs the process; the AI generates the raw material.

2. Agentic AI (The Doer)​

Agentic AI systems are proactive. They use prompts to pursue goals through a series of autonomous actions.

The Agentic Lifecycle:​

  1. Perceive: Look at the environment.
  2. Decide: Choose an action.
  3. Execute: Perform the action (e.g., browse a web page, query a database).
  4. Learn: Observe the output.
  5. Repeat: Loop until the goal is achieved.

Example: A personal shopping agent. Given "Buy me a new laptop," it hunts for availability, monitors prices, handles checkout, and coordinates delivery, only asking for help if stuck.

3. The Shared Foundation: LLMs​

Both approaches rely on Large Language Models (LLMs).

  • In Generative AI, the LLM is the creative engine.
  • In Agentic AI, the LLM provides Reasoning Capabilities (Chain of Thought).

The agent uses the LLM to generate an internal dialogue: "First I need to check the budget... then I need to find venues... then I need to check availability." This allows it to break down complex tasks into logical steps, similar to how humans solve problems.

4. The Future: Intelligent Collaboration​

The most powerful systems won't be purely generative or purely agentic. They will be Collaborators that know when to explore options (Generate) and when to commit to actions (Act).