Retrieval as the foundation
Enterprise search was positioned as the strategic foundation for AI. In 2024, I developed core RAG components and then evolved them into an agentic architecture integrated with enterprise search and company data.
Roche Diagnostics · Experience from delivery
From enterprise search to a global AI platform.
Two connected areas of work show how I bring together product strategy, technical architecture and team enablement.
01 / Service Copilot · January 2024–December 2025
An AI product for global service.
The starting point
Service knowledge sat in a complex knowledge landscape: around 4,000 products, roughly three million content assets, more than 20 content types and over 50 million content chunks. The task was to make this foundation usable for a scalable AI application.
My responsibility
I developed the vision, the initial prototype and the business case for the global AI program. As product owner and later technical product owner, I connected roadmap and architecture decisions with my own development work and close collaboration with the engineering team.
Learning from usage
A structured feedback system translated input from the service organization into prioritized requirements. Further development followed actual usage.
Enabling a team
I designed the initial recruiting process and played a key role in recruiting and onboarding 15 developers. Established responsibilities were handed over to successors.
The course correction: feedback needed its own process.
In the original project report, I also describe what did not work at first: the initial feedback approach did not scale. A dedicated feedback team and a systematic process then translated user input into prioritized requirements.
Context: the statement about the first approach comes from my project report. The Roche reference letter confirms that the structured feedback system was set up; it does not describe the earlier approach.
More than 10,000service professionals worldwide use the globally rolled-out AI platform.
Source: reference letter from Dirk Schäfer, Head of DSS, Roche Diagnostics GmbH, January 2026. The figures describe the program and its knowledge landscape.
02 / Enterprise search · from June 2022
Improving search quality measurably.
As Product Owner Search, Structure & Innovation, I worked on internal knowledge search at Roche Diagnostics. The starting points were user needs and the question of how to evaluate changes to the search system reliably.
From user interview to algorithm
I developed the core algorithm for the Azure search engine and dedicated tools for building and testing queries. Measurement scripts made the effect of changes on search results and on the performance of individual queries verifiable.
Feedback and adoption
User interviews, analytics and feedback loops connected technical development with the work of the content and product teams. In April 2023, I also contributed to the first internal ChatGPT prototype.
Source: reference letter from Andreas Bauer, GCS KLS Leader of Knowledge Management Solution, Roche Diagnostics GmbH. The letter itself is undated.
References
What Roche says.
What distinguishes Mr. Henze is not only his technical capability, but his ability to combine product vision, architectural clarity, execution strength, and organizational influence.
Hendrik consistently demonstrated rapid problem-solving skills, data-driven concept development, superb communication, and innovative thinking.
Full reference letters and a reference call are available on request.
The projects behind the references →What here could help your initiative?
The scale does not have to be the same. Questions about usable knowledge, clear architecture decisions and sustainable delivery come up in smaller organizations, too.