What are AI Agents

An AI agent is a system that receives information from its environment, makes decisions, performs actions, and uses the results of those actions to continue working toward a goal.

Although implementations and frameworks continue to evolve, several fundamental agent patterns are commonly used.

1. Simple Reflex Agent

A simple reflex agent selects an action based only on the current input.

Input → Rule → Action

It does not maintain memory of previous states.

Example: If a monitoring system detects that disk usage is above 90%, it triggers a cleanup action.

2. Model-Based Agent

A model-based agent maintains an internal representation of the environment.

Input → Update State → Decision → Action

This allows the agent to use previous information when deciding what to do next.

For example, an agent processing a dataset may remember which files have already been validated.

3. Goal-Based Agent

A goal-based agent selects actions according to a defined objective.

Instead of simply reacting to input, it evaluates:

Current state → Desired state → Possible actions

For example, if the goal is to prepare a dataset for training, the agent may determine that it needs to validate the schema, remove corrupted records, check duplicates, and generate a final report.

4. Utility-Based Agent

Sometimes several actions can achieve the same goal.

A utility-based agent evaluates the available options and selects the one expected to produce the best result.

Possible actions → Utility evaluation → Best action

Utility may represent cost, accuracy, latency, risk, quality, or a combination of several metrics.

5. Learning Agent

A learning agent improves its behavior based on experience or feedback.

A classical learning-agent architecture contains:

• Performance element  - selects actions.

• Learning element - improves the agent.

• Critic - evaluates the results.

• Problem generator - encourages exploration of alternative actions.

The important difference is that the agent's behavior is not completely fixed.

It can use feedback from previous results to improve future decisions.

Final Notes

Modern AI agents often combine several of these types. Understanding the basic patterns makes it easier to study more advanced agent architectures as they continue to evolve.