Agentic AI/ AI agents

There is different type of AI available, which can mainly be categorized into 3 categories

•        RAG based Chatbots
•        Tool Augmented Chatbots, mainly Automation.
•        Agentic Chatbots.

At Anthropic, all the variations are categorized as agentic systems, but an important architectural distinction is drawn between workflows and agents.

• Workflows are systems where LLMs and tools are orchestrated through predefined code paths.
• Agents, on the other hand, are systems where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks.

RAG based Chatbots

RAG chatbots are generative AI apps that combine retrieval and generation models to enable more accurate and relevant responses than traditional chatbots.

Example of a RAG based chatbot is a HR Assistant, Customer Support Chatbot etc.

 

Tool Augmented Chatbots (Automation)

Tool-augmented chatbots leverage external tools and APIs to enhance their capabilities beyond the limitations of their pre-trained knowledge. They not only provide relevant responses and but also can take an action.

Here are some examples of how they are used across different industries:

Human Resources – Leave Application: Chatbot can access the HR systems and help employees to apply for leave.

Health Care - Appointment Scheduling and Reminders: Chatbots can access appointment calendars and scheduling systems to allow patients to book appointments, receive reminders, and manage their healthcare schedule.

Travel & Hospitality - Trip Planning and Booking: Chatbots can access flight and hotel booking APIs to help users find deals, compare options, and book their travel arrangements.

Agentic Chatbots

Agentic AI, or AI agents, are software systems designed to autonomously perform tasks with minimal human intervention. They are goal-oriented and can interact with data and tools to achieve specific objectives. 

Essentially, they are digital workers that can sense, reason, and act based on programmed objectives, much like a human employee but operating in the digital realm. 

The main characteristic of Agentic AI.

•       Goal oriented planning – not giving tasks. Complex tasks which needs reasoning
•      Multi- step reasoning.
•       Autonomous Decision Making
•      Tools, Knowledge and Memory

The main components of an AI agents are: -

•       Brain (LLM’s) – Handles the reasoning, planning and language generation.
•       Memory – Enables the agent to remember past interactions.
•      Tools – Enables the agent to interact with the outside world.