End-to-End Development Workflow with AI Agents

The Used Car Platform is a central digital sales channel for Audi’s B2B used car trade. It includes, among other features, live auctions, a B2B marketplace, and AI-powered price predictions. Comsysto Reply has been supporting Audi in the further development of the platform since 2021. Together, we replaced the original legacy solution with a modern, AWS-based cloud architecture and have continuously expanded it into a scalable digital marketplace.

As the platform’s functionality grew, so did the demands on the development team. Operational tasks such as requirements analysis, code generation, quality assurance, and documentation increasingly tied up resources that were then lacking for the strategic further development of the platform. Together with Audi, Comsysto Reply therefore developed a customized multi-agent workflow that specifically integrates artificial intelligence into the software development lifecycle. The focus is not on a single assistant, but rather on the interaction of several specialized

‍

The Challenge

Creating More Room for Value-Adding Development

The Used Car Platform is continuously being expanded with new features and sales formats. At the same time, software development must meet the requirements of a business-critical platform, a complex system landscape, and clear quality and governance standards. For the development team, this meant that a significant portion of their time was spent on repetitive and time-consuming tasks. These include, among other things, the structured analysis of requirements, code development, quality assurance, and technical documentation.

The goal, therefore, was not to replace developers. Rather, the aim was to create a development model in which AI agents take on operational tasks, thereby freeing up more time for the team to focus on architecture, business decisions, and the development of new features.

‍

The Solution

Comsysto Reply developed a customized multi-agent workflow for the Used Car Platform based on GitHub Copilot. The solution supports the development process from requirements analysis through implementation to quality assurance and documentation.

The specialized agents work together to form a seamless end-to-end workflow:

1. Requirements Analysis and Planning

A specialized agent assists with the structured analysis and preparation of requirements. This ensures that business and technical information is consistently prepared for the subsequent development steps.

2. Code Generation

Based on the analyzed requirements, another agent assists in creating the necessary code logic. This relieves developers of repetitive implementation tasks, allowing them to focus more on solution design, architecture, and business decisions.

3. Quality Assurance and Documentation

Another agent assists with quality assurance and the automated creation of technical documentation. The results are then reviewed and validated by the development team and integrated into the ongoing development process.

‍

‍

The strength of this setup lies in the combination of artificial intelligence with the Used Car Platform’s specific system and domain expertise. The agents not only provide support through generic development patterns but are also tailored to the specific requirements of the platform and the development process. This enables them to deliver consistent results while taking into account Audi’s technical and organizational framework.

‍

People Remain at the Center

The multi-agent workflow is based on a human-in-the-loop model. The AI agents handle repetitive and time-consuming tasks, while developers remain responsible for review, validation, and final approval. The agents support the process but do not make final decisions on their own. As a result, quality, control, and responsibility remain firmly anchored within the development team.

At the same time, the developers’ focus shifts. Automated support for requirements analysis, code generation, quality assurance, and documentation frees up time for an earlier and more in-depth examination of architecture, solution design, and technical value. The resources thus freed up not only support faster implementation but also the deliberate and sustainable further development of the platform.

‍

The Result

The multi-agent workflow reduces routine tasks and helps development teams deliver new features faster and more efficiently. The time saved can be specifically invested in further developing the platform, creating new features, and responding to changing market demands in the B2B used-car trade. The result is a development workflow that does more than just speed up individual steps; it supports collaboration between developers and AI throughout the entire software development lifecycle.

For Audi, this creates a development model in which AI supports operational implementation, allowing teams to focus their resources more heavily on architecture, innovation, and business-critical features. With this project, Comsysto Reply demonstrates how long-standing platform expertise, cloud-native software development, and applied artificial intelligence can be combined into a seamless and practical agent-based development process.

‍