Accelerating AI-Powered Software Development: Hermes Scales AI Coding Adoption in Just Three Weeks
Hermes Germany GmbH is one of Germany’s leading logistics providers, offering comprehensive domestic and international delivery services. To maintain high service quality and customer satisfaction, the company continuously invests in the digitalization of its processes and IT landscape.
The project at a glance
- Higher developer productivity: Teams reported faster delivery cycles and improved development velocity.
- Improved code quality: AI agents identified long-standing defects that had previously gone unnoticed, contributing to measurable quality improvements.
- Sustainable capability building: Four development teams with a total of 20 engineers gained practical expertise in advanced AI-assisted development, including Context Engineering, the Model Context Protocol (MCP), agents, and subagents.
- Effective tool adoption: Teams shifted from traditional AI assistants to more powerful command-line-based tools such as Claude Code and Gemini, while less effective solutions were phased out.
Background
Existing Investments, Limited Adoption
Hermes had already invested in generative AI coding tools to explore their potential and improve developer productivity across Java-based services (Spring Boot, Quarkus, Micronaut), Vue applications, and Flutter projects.
Despite strong interest from developers, adoption remained limited. Teams lacked a structured approach and a deeper understanding of how to leverage AI effectively in real-world software engineering scenarios. As a result, the available tools were often used only superficially and failed to deliver their full value.
Unlocking Untapped Potential
It quickly became clear that providing access to AI tools alone was not enough. To achieve meaningful productivity gains, development teams needed practical guidance on how to integrate AI into complex microservices architectures and full-stack development workflows. To accelerate adoption and establish sustainable best practices, Hermes partnered with codecentric.
Solution
A Three-Week AI Sprint
Focusing on the Techniques That Matter
Learning Through Communities of Practice
A Three-Week AI Sprint
Together, Hermes and codecentric designed a focused three-week AI Sprint aimed at rapidly building both confidence and competence in AI-assisted software development. The program was built around three core components: an intensive kickoff workshop, ongoing Communities of Practice (CoPs) and a dedicated retrospective and planning session.
Focusing on the Techniques That Matter
The sprint began with a hands-on one-day workshop. Beyond covering the fundamentals of Large Language Models (LLMs) and prompt engineering, the training focused on advanced techniques that have the greatest impact on developer productivity:
- Context Engineering
- Model Context Protocol (MCP)
- Agents and subagents
- AI-supported software quality practices
Following a strongly practical approach, roughly half of the workshop was dedicated to hands-on exercises. Participants applied the concepts directly to their own projects and explored real-world use cases rather than purely theoretical examples.
Learning Through Communities of Practice
A critical success factor was the introduction of regular Communities of Practice (CoPs). In sessions similar to daily stand-ups, approximately 20 developers from four teams shared experiences, challenges, and lessons learned. This allowed experts to provide immediate support for specific issues—such as ineffective prompting strategies during migration projects—while also introducing participants to the latest developments in the rapidly evolving AI ecosystem. The format not only accelerated problem-solving but also fostered continuous learning and enthusiasm through visible progress and peer-to-peer exchange.
Result
Immediate Productivity Gains
The impact was visible almost immediately. Developers moved from occasional and largely experimental AI usage to the deliberate application of advanced AI engineering techniques. Teams consistently reported productivity improvements across a wide range of development tasks.
These observations were supported by measurable outcomes: development teams completed user stories significantly faster, and leadership observed clear improvements in delivery speed. At the same time, teams reassessed their tooling landscape and quickly adopted more effective solutions where necessary.
Tangible Improvements in Code Quality
The sprint also delivered immediate quality benefits. In one notable example, a team used an AI agent to identify a persistent pipeline defect within minutes—a problem that had remained unresolved for years. Experiences like this demonstrated that AI can serve not only as a productivity accelerator but also as a valuable partner in quality assurance and troubleshooting.
Building Long-Term AI Capability
The final sprint session focused on reflection, knowledge transfer, and next steps. Participants defined concrete action items, including the continuation of Communities of Practice and the launch of additional AI initiatives within their teams. This ensured that both the knowledge gained and the momentum created during the sprint would continue to generate value long after the engagement ended.
Looking Ahead: Enterprise-Wide AI Adoption
Following the success of the initial sprint, Hermes and codecentric are expanding the initiative across the organization. Four additional development teams are currently participating in the enablement program to establish a common foundation for AI-assisted software development across the company. At the same time, the focus is expanding beyond engineering. A dedicated Product AI Sprint is enabling Product Owners and Agile Coaches to integrate AI-powered workflows and agentic tools into their daily work, strengthening collaboration between business and technology teams.
Johannes Modersohn
Team Lead Software Development, Hermes Germany GmbH
The advanced prompting techniques, Context Engineering, and the focus on MCP, agents, and subagents were the real game changers. Even colleagues with prior experience discovered entirely new possibilities, and everyone was surprised by the quality of results coding agents can achieve when guided effectively. The sprint fully met our expectations—great work by the entire team.
Johannes Modersohn
Team Lead Software Development, Hermes Germany GmbH
Any questions about the project?
Ready to unlock the full potential of AI-powered software development? Get in touch with us. We’ll show you where AI can create real value for your development teams—and how to turn that potential into measurable results.
Service Lead GenAI
Service Lead GenAI
Further reference projects
Find out about other successful projects that we have completed with our clients. Perhaps you will find ideas here for a use case in your own organization.
Ferdinand Ade