What Exactly Is an AI “Agent” and How Is It Different from Regular AI?

There are two main types of AIs: generative AIs and agentic AIs. Generative AIs respond to user prompts, while agentic AIs, or AI agents, are designed to work independently without needing constant user input.

An AI Agent is basically a system that you can assign tasks to, and it takes care of them on its own. In the past, using LLMs meant you were mostly restricted to Q&A. Now, we have AI Agents that can handle more complex tasks like building a website. For instance, you might ask an AI Agent to review a big GitHub repo and suggest improvements, and it could even create a forked version of your code for testing—saving you a lot of time!

A good way to think about it is: LLMs exist within chat boxes, whereas Agents can function across the web or on your computer. It’s all about taking AI capabilities beyond just chat, letting them perform tasks that you could do with a computer or phone.

From what I've seen, 'Agents' can do a variety of tasks for you, such as ordering an Uber, booking accommodations, or even scheduling appointments—basically mimicking human interactions with apps. There are some AI tools that promise a lot but end up being underwhelming, like one I tried that didn't deliver as expected.

If you're looking to experiment, I'd suggest downloading VSCode and testing out both generative and agent modes. They operate quite differently. Agents interact with multiple inputs, and it's like they manage smaller AI tasks simultaneously. Generally, you'd consider a program an Agent if it utilizes at least a couple of sources, including an LLM.