Introduction

The terms agents and agentics are essential in the discussion of Artificial Intelligence (AI). While an agent refers to an individual, autonomous entity capable of perception and action, agentics is the field of study that explores the design, behavior, and systems involving these agents. This document clarifies these terms, provides examples, and defines key concepts to better understand agent-based systems.

Objective

This glossary aims to:

• Define core concepts related to AI agents and agentics.
• Provide examples to illustrate key terms.
• Serve as a reference for professionals and enthusiasts exploring agent-based systems.

Core Concepts of Agents

• Agent: An autonomous entity that perceives its environment through sensors and acts upon it through actuators.
  Example: A robot vacuum cleaner detects obstacles and adjusts its path to clean the floor.

• Autonomy: The capability of an agent to operate independently without human intervention.
  Example: Self-driving cars navigating roads without direct control.

• Environment: The context or surroundings in which an agent operates, either physical (e.g., a factory floor) or virtual (e.g., a stock market simulation).

• Perception: The process by which an agent gathers data about its environment through sensors.
  Example: A weather station using temperature and humidity sensors.

• Action/Actuator: Actions are how agents modify their environment; actuators are the mechanisms that perform these actions.
  Example: A drone delivering a package by adjusting its rotors.

• Goal: The specific objective an agent aims to achieve.
  Example: A chess-playing AI aims to checkmate the opponent.

• Rationality: The agent's ability to make decisions that maximize the likelihood of achieving its goal.

• Beliefs: Information or assumptions an agent holds about its environment.
  Example: A navigation system believing a certain road is open based on live data.

• Intentions: Goals or plans the agent actively commits to achieving.
  Example: Planning the shortest route to deliver goods.

• Capabilities: The set of actions an agent can perform.
  Example: A humanoid robot's ability to walk, speak, and lift objects.

Types of Agents

• Simple Reflex Agents: React directly to perceptions using condition-action rules.
  Example: A thermostat turning on heating when the temperature drops.

• Model-Based Reflex Agents: Use an internal model of the environment to handle incomplete information.
  Example: A drone updating its path based on GPS and obstacle data.

• Goal-Based Agents: Choose actions to achieve predefined goals.
  Example: A GPS system recalculating routes to reach a destination.

• Utility-Based Agents: Select actions that maximize a utility function representing desirability.
  Example: An e-commerce recommendation engine suggesting products to maximize customer satisfaction.

• Learning Agents: Adapt and improve performance over time by learning from experience.
  Example: Chatbots trained to provide better responses with user interactions.

• Software Agents (Softbots): Exist purely in software environments.
  Example: Virtual assistants like Siri or Alexa.

• Robotic Agents: Interact with the physical world through mechanical actuators.
  Example: Industrial robots assembling cars.

• Multi-Agent Systems (MAS): Systems comprising multiple agents interacting to solve problems.
  Example: Autonomous drones coordinating for disaster relief mapping.

Agent Architectures and Related Concepts

• Agent Architecture: Defines how an agent perceives, reasons, and acts.
  Example: Subsumption architecture in robotic agents prioritizes basic survival functions over complex reasoning.

• Agent Communication: Mechanisms by which agents share information.
  Example: Chatbots exchanging data via predefined protocols in customer service.

• Agent Coordination: Methods agents use to organize efforts to achieve shared goals.
  Example: A fleet of autonomous delivery robots avoiding collisions while working in the same area.

• Agent Negotiation: Interaction where agents try to reach mutually beneficial agreements.
  Example: Autonomous trading bots negotiating prices in a stock market simulation.

• Agent-Based Modeling (ABM): A technique for simulating the actions and interactions of agents to study complex systems.
  Example: Modeling traffic patterns to optimize city planning.

• Emergent Behavior: Unpredictable, complex behavior arising from simple agent interactions.
  Example: Ant colonies forming efficient foraging paths without centralized control.

Agentics: The Study of Agents

• Agentics: The field of study focused on the design, behavior, and interaction of agents and multi-agent systems.
  Example Application: Research into swarm intelligence for robotic systems.

• Distributed Artificial Intelligence (DAI): A subfield addressing how multiple agents work together to solve problems.
  Example: Coordinating autonomous vehicles in traffic systems.

• Agent Theory: Frameworks for understanding agent behavior and reasoning.
  Example: Formalizing rational decision-making for autonomous systems.

• Agent Programming Languages: Languages designed for building agent systems.
  Example: JADE (Java Agent Development Framework).

Differences Between AI Agents and Agentics IA

• Definition:

  • AI Agents: Autonomous entities that perceive their environment, make decisions, and take actions to achieve specific goals.
  • Agentics IA: The field of study focused on the design, analysis, and understanding of agents and multi-agent systems.

• Focus:

  • AI Agents: Practical implementation and functionality of individual agents or specific systems.
  • Agentics IA: Theoretical and practical frameworks that govern agent behavior and interactions within systems.

• Scope:

  • AI Agents: Limited to the behavior and performance of a single agent or defined group of agents.
  • Agentics IA: Broader in scope, covering coordination, negotiation, and emergent behaviors within multi-agent systems.

• Examples:

  • AI Agents: Self-driving cars, delivery drones, or customer service chatbots.
  • Agentics IA: Research on swarm intelligence, agent communication protocols, or distributed problem-solving systems.

• Key Question:

  • AI Agents: "What can this agent do to achieve its goal?"
  • Agentics IA: "How should agents be designed and interact within a system to solve complex problems?"

• Output:

  • AI Agents: Actions, decisions, or results produced in specific scenarios.
  • Agentics IA: Insights, methodologies, and models for building and analyzing agents and systems.

• Relation to AI Systems:

  • AI Agents: Individual components or entities within AI systems.
  • Agentics IA: Provides the theoretical and practical tools needed to design and analyze these agents effectively.

Automations with n8n or Make.com: Agents or Agentics?

Automations built with tools like n8n or Make.com, especially those leveraging AI, can be understood within the framework of agents and Agentics. Here's how:

What is an Automation in this Context?

An automation in tools like n8n or Make.com consists of workflows designed to execute specific tasks autonomously, often triggered by events and incorporating AI for decision-making or data processing.

When is it an Agent?

An automation qualifies as an agent if it meets these criteria:

• Autonomy: It runs independently after being configured, requiring minimal human intervention.
  Example: A workflow that processes incoming emails, classifies them using AI, and sends automated responses.
• Perception and Action: It perceives inputs (e.g., new data or events) and acts upon them (e.g., triggering a follow-up process).
• Goal-Oriented Behavior: It operates to achieve a defined objective, like optimizing data entry or responding to user queries.

Such automations are simple agents that execute predefined sequences of actions based on conditions.

Why Is It Not Agentics?

While automations are implementations of agent concepts, they are not Agentics for these reasons:

• Agentics is the study and design framework for creating agents and multi-agent systems, focusing on how agents interact, adapt, and solve problems collectively.
• Automations created in tools like n8n or Make.com are products of this field but do not encompass the theoretical or analytical aspects that define Agentics.

Advanced Cases: Moving Beyond Simple Agents

Automations can become more complex and approach multi-agent systems if they:

• Incorporate Learning: Use AI to adapt workflows dynamically based on data or patterns.
• Coordinate with Other Systems: Interact with multiple agents or external systems to achieve shared goals.
• Optimize Decisions: Utilize decision-making models that go beyond predefined rules, such as using utility-based functions to determine the best course of action.

In such cases, these automations reflect the principles studied in Agentics but remain implementations of individual or coordinated agents.

Summary

• Simple Automations: Agents that perform autonomous tasks based on predefined workflows.
• Agentics: The theoretical and practical foundation enabling the design of such automations and their integration into larger, intelligent systems.

By understanding these distinctions, you can better classify the role of AI-driven workflows in the context of agents and the broader discipline of Agentics.