At the core of an agent lies its mind (or brain), where critical thinking occurs by utilising reasoning, logic, memory, and other essential capabilities to achieve its goals. If the mind represents the LLM, then agents are self-driving or autonomous LLMs (Deslandes, 2025).
An agent is an autonomous application designed to reach a goal by observing and acting upon the world with available tools. It can function independently of human intervention, proactively planning and executing tasks to fulfil its objectives, employing reasoning, logic, and external information.
When we work directly with LLMs, we have a goal in mind. We craft prompts that are our requests and get a response. Then, it is a matter of tailoring the requests and checking the responses until they meet our intended goal. Prompt engineering is both the art and science of trying to get the best response with the least amount of effort. And in doing so, prompt engineering can easily force us to write like machines to get those better results.
Agents attempt to pull us away from the “voice of the machine” by converting arbitrary plain language into arbitrary digital commands executed by the LLM on our behalf. The introduction of agents enables us to speak to machines like anyone else. A universal translator for the masses to issue digital commands that would typically require knowledge worker professionals such as coders, doctors, lawyers, marketers, etc. In short, agents equip you with your digital team to get whatever you need to get done.
