Generative vs Agentic AI: Shaping the Future of AI Collaboration

Generative AI and Agentic AI

• Generative AI is primarily reactive: it responds to a prompt by generating text, images, code, audio, or other content. It relies on patterns learned from large datasets and generally stops once it has produced an output. A human typically reviews, selects, and refines that output.
• Agentic AI is proactive and goal-oriented. After receiving a goal, it can repeatedly perceive its environment, choose and execute actions, and learn from the results, with limited human intervention. It is suited to tasks involving ongoing management and multiple steps.
• Shared technology: Both approaches may use large language models (LLMs). LLMs power many chatbots and can provide reasoning for agents; image and audio generation may use other models, such as diffusion models.
• Reasoning and example applications: Agents can use chain-of-thought reasoning to break complex tasks into smaller steps. For example:
  • A shopping agent might search across platforms, monitor prices, complete checkout, and coordinate delivery.
  • A conference-planning agent might clarify requirements, find suitable venues, and check availability.
• Outlook: Future AI systems may combine both approaches—generating options when exploration is useful and taking action when it is time to pursue a goal.

Reviews, Guides, or Tutorials

No product review or step-by-step tutorial is presented. The video offers an explanatory overview with practical examples.

Main Speaker and Source

Presenter Martin Keen; IBM Technology.