Role-based AI Agents

AI Agent

Definition

An AI Agent is a software entity that operates autonomously as an independent system and interacts with tools and systems.

• Autonomous: Capable of making decisions without human intervention
• Independent: Operates as a self-contained system
• Interactive: Communicates with external systems and environments

Types of Agents

Although there is no universally established taxonomy, AI Agents can be broadly categorized by their functional orientation:

• Collaborative Agents: Perform specific roles in task environments, collaborating with humans to increase productivity (e.g., AI developers)
• Supportive Agents: Act as assistants that provide support across daily activities, similar to personal assistants (e.g., AI secretary)
• Delegated Agents: Operate as proxies for humans, handling tasks with low risk or low decision-making complexity (e.g., robo-advisors)

Components of an Agent

AI Agents are designed to mirror the structural and functional characteristics of humans. Their components include:

1. Persona

The personality and identity that the agent expresses during interaction.

Attribute	Description	Examples
Name	Unique identifier	Chloe, Siri, Jarvis
Personality Traits	Behavioral tendencies	Friendly, humorous, serious
Emotional Expression	Interactional emotional feedback	Comforting, encouraging
Conversational Style	Manner of speech	Formal, casual
Persona Type	Role identity	Support agent, consultant, companion

2. Ability

The overall capacity to perform tasks or solve problems.

Attribute	Description	Examples
Cognitive Abilities	Understanding and processing	Natural language understanding, pattern recognition
Reasoning	Problem-solving logic	Causal reasoning, decision-making
Interaction Capabilities	Communication modes	Voice, text, multimodal
Learning Capabilities	Learning new knowledge	Reinforcement learning, supervised learning
Adaptability	Contextual responsiveness	Personalization, context-awareness

3. Skill

Functional capabilities for performing specific tasks.

Attribute	Description	Examples
Skill Name	Name of the function	Weather inquiry, scheduling
Skill Type	Category of function	Q&A, recommendations
Input Format	Required input type	Text, voice, image
Output Format	Output type	Text, image, voice
Accuracy	Performance metric	95% image classification accuracy

4. Knowledge

Structured and unstructured information that the agent possesses.

Attribute	Description	Examples
Domain	Specialized areas	Healthcare, finance, education
Source	Information origin	Wikipedia, databases, papers
Currency	Update recency	As of March 2024
Depth of Knowledge	Expertise level	General vs. expert knowledge
Representation	Data structure	Ontology, knowledge graphs

Role-based AI Agents

Role

Future AI agents will be assigned distinct roles, much like humans. Based on these roles, they will be authorized to perform missions and operate within defined boundaries. Rather than R&R (Roles and Responsibilities), agents will follow the concept of R&A (Roles and Authorization).

Mission

A mission refers to a goal or objective the agent must accomplish. Missions vary in form and are contextually assigned based on the agent's role, guiding the execution of tasks.

Workflow

To achieve a mission, agents construct workflows—sequences of tasks designed to fulfill a goal. These workflows may involve internal models or external APIs. For example:

• Creating a Meeting Schedule Workflow

  

  1. Search for a suitable venue based on meeting purpose and constraints using a map API
     • Purpose: "Book Club Meeting"
     • Attendees: 5 people
     • Availability: Weekend
  2. Reserve the selected venue via booking API
  3. Register the confirmed event on a calendar via calendar API
  4. Send notifications to attendees using a messenger API

Task

A task is the smallest unit of work an agent can perform. Tasks are defined by specifications and can number in the thousands depending on mission complexity.

Example: Role-based Agent

• Role: Home management agent
• Mission:
  • Maintain home cleanliness
  • Optimize indoor environment (temperature, humidity, air quality)
• Tasks:
  • Cleaning
  • Waste sorting
  • Operating air purifier
  • Adjusting temperature
  • Running dehumidifier
  • Doing laundry