In my last post, Demystifying AI Agents: Separating Fact from Fiction, we explored the difference between a simple AI assistant and a truly strategic AI Agent. We saw how the team at the luxury car company, Build Your Car (BYC), could use agentic patterns to tackle big-picture goals.
That was the view from inside the company. But to truly understand the power of this technology, we need to see it through the eyes of the person it’s meant to serve: the customer.
Let’s meet Alex.
Alex has been saving up and is finally ready to buy his dream car from BYC. He’s excited about their newer series, the rugged Terra and the sleek Luxe. But his excitement is quickly turning into confusion.
His journey starts on the BYC website. He’s trying to compare the two models, but the information is spread across multiple pages. What are his options? He could try the website's chatbot, but he knows it will likely just give him links he’s already seen. He could fill out a contact form, but he doesn’t want his phone to start buzzing with calls from sales agents when he’s just trying to gather information. All he wants is a simple, smart way to figure out which car is right for him.
This is a classic problem. For customers like Alex, the buying process is often disconnected and frustrating. For companies like BYC, they are managing numerous channels and trying their best to help, but they risk losing potential customers like Alex who simply give up.
This is where Team Symphony, envisioned a new path forward. We imagined an intelligent, unified platform that could solve this exact problem an Agentic Platform we called OmniConnect. The recent IBM AWS Hackathon 2025 gave us the perfect opportunity to bring this vision to life by building our first Proof of Concept.
The Solution: OmniConnect, BYC’s AI Agent Concierge
OmniConnect isn't just a chatbot. It's an AI Agent Concierge built using the AWS Agentic Framework. Think of it as BYC’s single most helpful employee, available 24/7 to provide a seamless, personalized car-buying experience. It’s the practical application of the agentic patterns we discussed before: a multi-agent system where a conductor agent works with specialists that can use tools to get things done.
Under the Hood: A Look at the AWS Agentic Architecture
Building an agent this capable requires more than clever code, it takes a solid foundation and a team that works in perfect sync. Behind OmniConnect’s smooth customer experience lies a carefully orchestrated mix of AWS services, each playing a role in bringing the agents to life.
The Foundation: A Shared Brain and Memory
Every capable agent needs two essentials: a sharp mind to reason and a dependable memory to recall the right facts at the right time.
§ The Brain
For OmniConnect, that mind is Amazon Bedrock, powered by the Claude 3.5 Sonnet model. This Brain gives every agent the ability to understand user requests, think through the necessary steps, and respond in a way that feels natural, accurate, and human-like.
§ The Memory
The Memory is a Bedrock Knowledge Base backed by Amazon OpenSearch Service. This is where all approved, trusted company knowledge lives — from BYC’s official eBrochures to pricing sheets — processed and indexed by the Amazon Titan G1 Text model. Whenever an agent answers a question, it retrieves the most relevant, up-to-date information from this Memory, ensuring every response is accurate, consistent, and brand-approved.
The Multi-Agent Team
OmniConnect isn’t one single agent; it’s a team of specialists designed for collaboration.
1. The Conductor: The Orchestration Agent
Alex’s journey begins with a simple question typed into the chat. The Orchestration Agent is the first to greet him, quietly listening and figuring out what he needs. When Alex asks to compare two car models, the Conductor doesn’t try to answer on its own - instead, it hands the baton to the Sales Inquiry Agent. Later, when Alex wants a test drive, it knows that’s a job for the Test Drive Booking Agent. Think of it as the traffic controller, keeping conversations moving smoothly and making sure each request reaches the right hands.
2. The Specialist: The Sales Inquiry Agent
The moment the Conductor makes the handoff, the Sales Inquiry Agent steps in like the ultimate product expert. Using the shared Knowledge Base, it crafts a detailed side-by-side comparison of the Terra and Luxe models - features, specs, and pricing, all neatly laid out. But this agent isn’t just about talking; it can act. If Alex shows real interest, it taps into its Action Group to call an AWS Lambda function that creates a formal sales lead in DynamoDB, handing Alex a unique Inquiry ID - a tangible next step toward owning his dream car.
3. The Specialist: The Test Drive Booking Agent
Once Alex decides it’s time to get behind the wheel, the Orchestration Agent calls in the Test Drive Booking Agent. This agent begins by discreetly retrieving Alex’s pincode from DynamoDB, with all personal data masked and encrypted for privacy. It then reaches out through its Action Group to locate the nearest BYC dealerships, checks live vehicle availability and test drive slots, and returns with a tailored set of options. When Alex makes his choice, the agent locks in the booking and hands over a confirmation ID - his golden ticket to the driver’s seat.
4. The Silent Partners: AWS Services Behind the Scenes
While Alex chats away, a network of AWS services quietly keeps the whole system running. Amazon Bedrock fuels the reasoning, Amazon OpenSearch organizes the knowledge, DynamoDB stores leads and customer data, Lambda functions handle specialized actions, and secure APIs ensure that every piece of information moves exactly where it’s needed - fast, accurate, and safe.
The Result: A Journey Transformed
With OmniConnect, Alex’s entire experience changes. In a single, easy conversation, he gets a detailed comparison of the Terra and Luxe, receives a personalized recommendation for the Terra, and seamlessly books a test drive at a dealership near him for the next day. And if he wants to take it further, the agent can even create a follow-up appointment with the Sales Manager to discuss purchasing options.
No frustration. No annoying calls. Just a helpful and efficient process that gets him one step closer to his dream car. This seamless experience for Alex translates directly into powerful results for BYC. They can now offer a premium, personalized experience at scale, leading to happier customers and a more effective sales team. We project this can lead to a ~20% improvement in customer engagement and a ~25% boost in conversion rates, all while reducing operational costs.
Note (Learning Resource): AWS provides a sample implementation of a retail agent using Amazon Bedrock, which you can explore for hands-on learning and experimentation.
Amazon Bedrock Retail Agent – GitHub Repository
Conclusion
Our journey with OmniConnect, from a simple idea to a functioning AI agent, has been a powerful lesson in the practical application of agentic AI. For customers like Alex, this technology means replacing frustration with a guided, personalized experience. For businesses like BYC, it represents a fundamental shift towards more efficient, intelligent, and customer-centric operations.
The AWS Agentic Framework, with services like Amazon Bedrock at its core, provides not just the building blocks, but a robust and scalable foundation for turning these ambitious concepts into reality. As we continue to push the boundaries of what's possible, it’s clear that the future lies not in standalone models, but in these orchestrated systems of intelligent agents working together. This project was our first step, and it has solidified our belief that the most exciting part of the agentic journey is what lies ahead.
I hope this write-up has been insightful and informative, and may inspire you to start yours.
Thank you for reading! Your feedback and suggestions are welcome.
@Indranil Dey - Data Architect
@Krishna Choudhury - Solution Architect
IBM India, Kolkata