Riedel Communications GmbH.

Data sat across multiple ERP systems, each department saw only its own slice, and 70% of SAP master data records carried issues that flowed into every report and process downstream.

Built the BI layer across the ERP landscape, covering the full order-to-delivery chain, and brought SAP master data from a 70% error rate to 2% within a year through systematic audits and governance rules at the point of entry. Product management, operations and procurement now work from the same reliable data, which has saved over €500K, most visibly in procurement.

Seven functions, from product management to controlling, each held a different picture of where a product stood in its lifecycle. Decisions waited on email chains and suffered from version confusion.

Built a single lifecycle view now shared across all seven functions. Decisions move faster, and the errors that came from everyone working off a different version of the truth have largely disappeared.

Creating a pricelist took over 3 months in Excel, with seven departments working in silos and no structured way to reflect market changes, such as trade taxes, in the pricing calculations.

Owned the change end to end: the business case, alignment across all seven departments, the technical architecture, delivery with IT and data science, and the new operating process around the tool. Role-based workspaces in Azure AD let each department manage its own part, with changes flowing transparently through a defined cycle. A pricelist now takes under a month, addendums a few days, and manual uploads are gone.

The company saw potential in generative AI but had no platform, no roadmap, no prioritised use cases, and no framework for delivering them.

We started with one project, the Riedel Internal Assistant. I led it as the test case for which platform fit Riedel, and together we chose Copilot Studio inside M365 over a standalone app. The assistant answers policy, SOP and training questions in seconds instead of days, delivered in December 2025. More importantly, the project gave the company its GenAI foundation: the PRD format, the delivery process, and the first governance path for GDPR and the EU AI Act, all of which the later products reused.

One product proved the value, but it also showed the limit. AI use was spreading across the company faster than anyone could check it, most of it unreviewed. The EU AI Act was bringing new duties on training and high-risk systems. And one team could not build for every department. Blocking people was not an option; they needed a safe, fast path.

I set up the AI Enablement Team and lead it. Its job is to coach, not to build for everyone: departments own their ideas and results, we give them a fast way to ask, safe tools, clear rules and help when they need it. I designed the operating model with the team, the AI Front Door that turns most requests around in days, the approved tool library, the risk checks, the training that meets the EU AI Act's AI-literacy duty, and I own the link to the AI Committee and management. Alongside the framework, I led the two products the team built end to end: the Multi-Agent IT Helpdesk on AWS Bedrock, serving 1,500 employees, live since June 2026 and deflecting around 15% of L1 tickets since go-live; and the Tender & Product Bot on Azure AI Foundry, which answers spec and pricing questions instantly and, with human-in-the-loop tender creation, brought tender time from a month to two weeks. Both products run as production systems in their own right; the enablement framework is what lets other departments bring their own.

Shipping one product is an achievement; shipping repeatedly needs structure. In the EU, AI without a clear governance path does not reach production at all.

Built the governance layer covering GDPR compliance, EU AI Act risk classification and works council alignment, and established GitHub, CI/CD and structured project management as the team's delivery standard. The result is AI that clears legal review, delivered by a process the team can repeat.