What Are AI Agents?
Greg Asquith
An AI Agent is a computer program that is designed to perform tasks with minimal human input to achieve a certain goal or target.
The main point of difference between Agents and other AI tools is that Agents can decide what they are going to do, in which order, to complete their tasks.
They also often have a set of "tools" they can use to interact with other systems (e.g. Using an API to acces an ad platform).
Agents are to AI systems as 'apps' are to mobile operating systems - a foundational unit of capability which is useful in isolation, and makes the whole system more capable when combined together.
Similar systems, like ChatGPT, Gemini or Claude, rely only on their training to provide answers to users. Agents have the same training plus the ability to use other tools to achieve their goals.
A good metaphor is schoolwork -
Chatbots answer questions like they are doing an exam - they have only a pen and paper and have to rely on already knowing what they want to write. This is why sometimes they use infotmation that is not up to date or make up facts while trying to answer questions (we've all done it!).
Agents work like they are doing coursework/research - they can have access to the internet, books and other tools to help them. They can even review their own work from a different point of view (like a human taking a day off and coming back with a fresh mind).
Agents can appear like chatbots, with a simple text-based user interface, run entirely in the background or somewhere in-between, based on other types of input. They're very flexible, which means their real value is found when they are customised to your exact use case.