What is an Agent?

Artificial intelligence (AI) agents represent a significant evolution in software systems, leveraging AI capabilities to pursue objectives and accomplish tasks on behalf of users. These sophisticated systems exhibit reasoning, planning, memory, and a notable degree of autonomy, enabling them to make decisions, learn from experience, and adapt to changing circumstances.

In essence, AI agents are autonomous intelligent entities that perform specific tasks without the need for continuous human intervention. At their core, AI agents are characterised by a set of key capabilities that enable their autonomous operation. These include the ability to perceive their environment through sensors or data inputs, to reason about the information they gather, to take actions based on their understanding, and to learn from feedback and experience to improve their performance over time. This allows them to pursue goals, complete tasks, and adapt to new information in real time.

Their capacity to make rational decisions based on their perceptions and available data is crucial for achieving optimal results. Moreover, AI agents can understand and interpret user queries expressed in natural language with minimal human guidance. The fundamental principles that define these agents often include autonomy, a clear orientation towards achieving specific goals, and the ability to adapt to dynamic situations.

In terms of implementation, an AI agent looks like this:

env = Environment()
tools = Tools(env)
system_prompt = "Goals, constraints, and how to act"

while True:
    action = llm.call(system_prompt + env.state)
    env.state = tools.run(action)

Agentic AI represents the overarching concept of artificial intelligence systems that can act independently to achieve specific goals. Within this framework, AI agents serve as the individual components that execute particular tasks. Essentially, agentic AI is the underlying technology that empowers AI agents to function autonomously without continuous human oversight. These agentic systems often comprise multiple AI agents that work in concert, leveraging the capabilities of Large Language Models (LLMs) to tackle complex problems.

TL;DR

AI agents use LLMs to decide their best paths of actions to complete tasks, without requiring human intervention.