Multi-Agent System and Autonomous Agents - Next Frontier of Generative AI

We are transitioning from an era of knowledge-oriented, general AI-powered tools such as chatbots designed for responding to inquiries and creating answers and content - to a New Phase where AI Agents and Agentic workflows leverage foundational models to carry out iterative, multi-stage processes within the digital operation landscape.

Technology Moving Towards Autonomous Operations with Multi-agent or Agentic Systems. Multi-agents/Agentic Systems are digital systems that can independently interact in a dynamic world.

Intelligent or AI agents are broadly designed as autonomous problem solvers with the ability to understand their environment, follow decision model processes, and take action on the environment

Agentic systems traditionally have been difficult to implement, requiring laborious, rule-based programming or highly specific training in machine-learning models. With large language models (LLMs) and Small Language Models along with Domain Specific Models and Language Action models, Now Agentic systems and AI Agents are capable of utilising natural language to perform a variety of intricate tasks in diverse environments, showcasing notable potential.

Agentic Workflows decompose complex tasks into subtasks to be performed by independent LLM/SLM agents or as skilled virtual coworkers or Virtual assistants, collaborating with humans seamlessly and naturally.

A Multi-Agent System is a distributed system consisting of multiple decision-making agents that interact and act autonomously to achieve common and conflicting goals.

As per the Alan Turing Institute,

• How to design Multi-Agents systems to incentivise certain behaviours in agents.
• How to design algorithms enabling one or more agents to achieve specified goals in a Multi-Agents system.
• How information is communicated and propagated among agents
• How norms, conventions and roles may emerge in Multi-Agents systems.

A vast array of applications can be addressed using Multi-Agents systems methodologies, including autonomous driving, multi-robot factories, automated trading, commercial games, automated tutoring, etc.

Potential Multi-Agents Use Cases -

1. Housing Loan Application and Approval Process.

2. Project Management and Risk Assessment

3. Code Reviews, Documentation and Modernisation

4. Multi-Language Technical Documentation for the Products

5. Marketing Campaign

6. Personalized outreach with Data Enrichment

OpenAgents: AN OPEN PLATFORM FOR LANGUAGE AGENTS IN THE WILD

McKinsey Shared the Flow of Multi-Agent System

Microsoft AI Research Proposed Architectures

What We Building Compound AI Systems - Agentic Systems for Enterprise Data