Agents of Change: Multi-Agent Systems in Retail for Efficiency and Innovation

Introduction to Multi-Agent Systems (MAS)

Multi-agent approaches to AI applications are gaining prominence as a transformative paradigm. These approaches leverage the collaborative capabilities of multiple foundation model-based agents to tackle increasingly complex tasks with greater efficiency and effectiveness.

Multi-agent systems bring together diverse AI agents, each with specialized skills and knowledge, to work in concert towards a common goal. This collaborative intelligence enables the decomposition of intricate problems into manageable sub-tasks, which individual agents can address using their unique strengths. By integrating their efforts, these agents can achieve solutions that are more robust and comprehensive than those derived from isolated AI models.

Agents possess skills that you define. These skills determine their abilities and knowledge. For instance, an agent might excel at sentiment analysis, entity recognition, or summarization.

The Agents also have models associated with them. Fine-tuning models for specific topics is crucial. For example, fine-tuning a language model on retail-related text can create an agent specialized in retail conversations.

Chain of Thought Prompting

Multi-agent AI systems involve the collaboration of multiple AI agents, each with specialized capabilities, to autonomously handle complex tasks. These systems leverage advanced prompt engineering techniques like Chain of Thought prompting, which guides the AI to “think aloud” and iterate through its reasoning process step by step. This approach is particularly effective for multi-step reasoning tasks, allowing each agent to contribute to a segment of the problem-solving process. By breaking down tasks and employing iterative action planning and decision-making, multi-agent AI systems can efficiently synthesize information and generate comprehensive responses to user queries

An Example in Retail

Here is a possible example of how orchestration could work between agents that are specialized in specific tasks:

Some Multi Agent Systems

The number of multi-agent AI systems is rapidly expanding, with several players entering the field. Among them are Autogen & Autogen Studio (from Microsoft), Crew AI, langGraph, llama-agents, to name a few. These systems bring diverse capabilities and applications, making MAS an exciting area of innovation!

• Autogen: AutoGen is a framework that enables development of LLM applications using multiple agents that can converse with each other to solve tasks.

• Autogen Studio: A low-code interface for rapidly prototyping multi-agent AI solutions, built on the Autogen framework.

• Crew AI: A framework for orchestrating collaborative, role-playing AI agents to tackle complex tasks.

• Llama Agents: An open-source framework for building and deploying multi-agent AI systems as production microservices.

• LanGraph: A library for constructing stateful, multi-actor applications with LLMs, enabling cycles, controllability, and persistence.

LOB Systems using Agentic AI

Agentic AI Systems offer a transformative approach to building line-of-business applications by leveraging specialized agents with distinct skills and fine-tuned models. These agents can be orchestrated using Multi-Agent Platforms like Autogen, creating a dynamic and flexible system akin to microservices in traditional software architecture.

Experimenting with Autogen

I’m excited about the potential of Agentic Systems and have been actively exploring both Autogen and Autogen Studio. For a practical, hands-on learning experience with Autogen, I suggest checking out the course offered by Deeplearning.ai on AI Agentic Design Patterns, which you can find at their website under the short courses section. If you’re looking for a practical guide to Autogen Studio, Matthew Berman’s “AutoGen Studio 2.0 Tutorial” is an excellent resource. It was an absolutely fun experience for me to follow the step-by-step instructions on how to create agents that can extract transcripts from YouTube videos and transform them into engaging blog posts. I recommend setting up Autogen Studio with Docker to ensure that the Execution Policy on your OS doesn't prevent it from running code. This blogpost was very useful for helping resolve the execution issues I was having: https://blog.finxter.com/how-to-set-up-autogen-studio-with-docker/

The Startup Ecosystem

Some of the Startups that are building solutions using Agentic Systems include AUI™ (Augmented Intelligence) , JoopiterX . I would love to hear from startups building solutions for retail using Agentic systems using Microsoft Azure.

Real World Examples

Agentic AI Systems have been successfully integrated into the Tabor AI platform for streamlining the annual health plan selection for Medicare brokers. Traditionally, a labor-intensive task that required several hours of human research is now efficiently handled by AI agents in just 5 to 10 minutes. This innovation maintains the high standards of quality and accuracy necessary for the 65 million seniors in the U.S. who undergo this process yearly. Here are the details: Tabor - AI for Medicare

Case Study: How Tabor Leverages Generative AI for Medicare Plan Selection (skypoint.ai)

While AI applications in retail may not demand such precision, the deployment of AI in heavily regulated sectors like healthcare, finance, and government services presents unique challenges, which Skypoint AI platform adeptly addresses.