Case study

ZEISS Microscopy and Endava Explore the Impact of AI-Based Tools in Complex Software Modernisation

Healthcare and Life Sciences
Hand interacting with digital AI interface symbolizing software modernisation through artificial intelligence.

Challenge

Over time, ZEISS Microscopy’s ZEN platform grew to more than four million lines of code and involved 100+ developers, increasing complexity. Continuous modularisation was essential to improve maintainability and accelerate development, but manual analysis and refactoring required significant effort.

Outcome

Using Compass, our internally developed agentic AI tool, we supported ZEISS in a structured pilot to identify and validate refactoring opportunities within complex parts of the ZEN platform. The work showed how AI can support refactoring activities and productivity gains.

Gen AI is rapidly transforming how organisations modernise and manage complex software systems. Yet, tangible evidence of its impact is often scarce. In a joint pilot with ZEISS Microscopy, we evaluated the use of AI-based tooling to refactor large, complex classes of the ZEN platform. 

 

For the pilot we used Compass, our internally developed agentic AI tool designed to support large-scale modernisation projects. While Compass was the primary driver in this initiative, the learnings also apply to other AI-based tools, such as Microsoft Copilot or GitHub Copilot, demonstrating both the opportunities and the limitations of AI in the software development life cycle (SDLC). 

Scientist analyzing cell images on dual monitors in a high-tech laboratory

Understanding a complex system and building a new approach

The ZEN platform is the core software solution on most imaging systems from ZEISS, supporting both routine workflows and advanced research experiments – from image acquisition and processing to visualisation and analysis. Over many years, ZEN has grown to more than four million lines of code and involves over 100 developers. This scale has increased complexity and made some large classes particularly challenging to manage. 

 

Continuous modularisation and simplification of the platform are essential to improve testability and maintainability, and to further accelerate the pace of feature development.  However, this is a high-effort and time-consuming undertaking. 

Developers reviewing code on laptop screen in modern workspace

Aiming to reduce code complexity while reducing manual refactoring overhead, Compass was used to help the team systematically identify refactoring opportunities within large classes. Over the course of five months, we focused on five use cases, comparing manually implemented solutions with AI-identified proposals and validating results with ZEISS SMEs. From the outset, we defined a structured approach and modernisation plan, covering the analysis, execution and validation phases, which proved critical to success. 

 

The project was supported by Microsoft. Compass was deployed in ZEISS’s Azure environment and integrated with the Azure OpenAI service. 

 

Now, with the pilot successfully completed, we are eager to share our learnings with other teams across ZEISS who may be facing similar challenges in software modernisation and exploring AI-driven efficiency. 

Professional presenting a strategy diagram on a whiteboard during a planning session

Key learnings

  • High-value strengths: Compass proved particularly effective in high-level code analysis, understanding business concepts, mapping workflows and dependencies, producing actionable backlogs, and creating step-by-step refactoring templates. 
  • Productivity gains: The use of AI led to measurable acceleration, including up to 5x faster issue detection, 5x backlog creation and 2x faster analysis, knowledge transfer and creation of solution proposals. 
  • Governance: To realise these productivity gains, a structured and guided approach is essential. While refactoring is an iterative process, establishing a clear modernisation plan before diving into detailed analysis was key to success. 
Scientist analyzing cell images on dual monitors in a high-tech laboratory

Accommodating for risk

Like any AI solution, Compass comes with boundaries. Outputs must be validated, and context carefully set to ensure reliable results. Compass is not designed to write complex code or to map low-level dependencies. 

 

Practical recommendations 

  • Start with high-level prompts, then drill down gradually 
  • Request fact-based analysis to reduce assumptions 
  • Use iterative follow-ups for refinement 
  • Fact-check and validate outputs with SMEs 
Close-up of screen showing data with highlighted text for development use

Positive impacts unlocked and the path ahead

The pilot clearly showed that AI is a powerful enabler for software modernisation when applied carefully. It delivered measurable productivity improvements throughout the SDLC, helping the team identify refactoring opportunities, reduce technical debt, accelerate development and improve maintainability of the ZEN platform. 

 

Building on this success, Endava and ZEISS Microscopy are discussing a scalable, repeatable modernisation process to systematically refactor selected high-impact areas of the ZEN platform. The approach suggests AI-assisted ‘refactoring pods’ combined with ZEISS SMEs to accelerate development while reducing manual refactoring overhead and ensuring high quality output. 

Interesting? We love when people share.

We can help you reach new heights

Let's connect