Client Overview

AMP GRUP is a leading player in the logistics and supply chain industry in Romania, renowned for its commitment to innovation and operational excellence. With a diverse portfolio and a forward-thinking approach, AMP GRUP continually seeks new ways to empower its teams and deliver value to its clients.

Business Challenge

AMP GRUP's sales team faced the challenge of accessing real-time, actionable data from their enterprise ERP system while on the move. Manual data retrieval processes slowed down sales cycles and limited the team's ability to provide timely, data-driven responses to client inquiries. The company needed a solution that would seamlessly integrate with existing systems and empower the sales force with instant insights.

Project Overview

To address these challenges, AMP GRUP partnered with our team to design and implement an AI agentic chatbot, purpose-built to support the sales team. The project aimed to leverage the power of Google Cloud and multi-agent architectures to create a digital co-sales member—an intelligent assistant capable of connecting to the enterprise ERP, retrieving relevant data, and providing contextual support to sales representatives in real time.

Solution

The solution centered on deploying a secure, scalable chatbot platform on Google Cloud. Key features included:

• Integration with AMP GRUP's ERP system for seamless data access

• Multi-agent orchestration to handle complex sales queries and workflows

• Natural language interface for intuitive user experience

• Role-based access and data security controls

• Real-time analytics and reporting for sales performance

The chatbot acts as a digital co-sales member, providing instant answers, surfacing relevant documents, and proactively suggesting next steps based on sales context.

High Level Design

A Look Under the Hood: Our Google Cloud Solution

When we set out to build this solution, our goal was to create something truly integrated and intelligent, leveraging the best of Google Cloud. I wanted to give you a quick overview of how we pulled it off.

The Building Blocks

Cloud Run for the Core: We chose Cloud Run as our workhorse for hosting. It's fantastic because it handles both our User Interface (UI) containers and all the backend functionality. The auto-scaling and high availability features mean we don't have to worry about managing servers, which really frees us up to focus on development.

Intelligent Agents with Vertex AI: This is where things get really interesting. We've deployed sophisticated AI agents using Vertex AI with Agent Builder. These agents are designed to be the bridge between our application and our internal systems, connecting directly to our ERP system, handling emails, and talking to other critical applications.

For the brains of these agents, we're leveraging Google's Gemini models. For high-volume, quick responses, we're using Gemini 2.5 Flash, which is great for its speed and efficiency. When we need deeper analysis, complex reasoning, or multi-step problem-solving, Gemini 2.5 Pro steps in. To ensure these agents can securely and efficiently interact with the ERP system, we've implemented the Python ADK (Agent Development Kit) and our custom A2A (Agent-to-Agent) frameworks. Communication with the ERP system itself is handled via the MCP (Message Communication Protocol), which provides a reliable and secure channel.

Keeping an Eye on Things with Cloud Monitoring: To make sure everything runs smoothly, we're using Cloud Monitoring. It gives us a comprehensive view of our operations, helping us keep tabs on performance, identify any issues quickly, and ensure overall system health.

Cloud Storage for Our Data Lake: For all our unstructured data – things like documents, images, and other large files – we're relying on Cloud Storage. It's incredibly scalable and durable, giving us a robust place to keep all that diverse information.

Vertex AI Vector Search for Deep Insights: Finally, to really power our AI agents' understanding, we're storing our graph embeddings datasets in Vertex AI Vector Search. These embeddings are essentially numerical representations of complex relationships within our data. We're generating them using the gemini-embedding-001 model, which is Google's latest state-of-the-art text embedding model. This allows our agents to perform incredibly efficient similarity searches and gain deeper insights into the connections within our data.

Results

The implementation of the AI agentic chatbot has transformed AMP GRUP's sales operations, delivering significant improvements in efficiency and customer service:

The chatbot has quickly become an indispensable member of the sales team, driving both efficiency and customer satisfaction.

Future Vision

Building on this success, AMP GRUP plans to expand the chatbot's capabilities to other departments and integrate additional data sources. The vision is to create a unified, AI-powered digital workforce that supports every aspect of the business, from sales and customer service to logistics and operations.