Körber accelerates GxP operations with trusted AI leveraging Microsoft Azure

Read the story

Bringing trusted AI into life sciences

For decades, pharmaceutical manufacturers have faced a difficult balancing act: modernizing production operations while maintaining the rigorous compliance standards required to protect patient safety.

Körber, a global technology group with approximately 13,000 employees across more than 100 locations worldwide, has spent years helping pharmaceutical manufacturers digitalize and manage production through its PAS-X Manufacturing Execution System (MES), developed by its BU Software. Today, the company supports approximately 1,500 active PAS-X MES systems across pharma, biotech, and cell & gene therapy environments. That installed base underpins a meaningful share of global patient care: Körber reports that PAS-X MES supports roughly a quarter of the global pharma MES market.

According to the Boston Consulting Group (Market Study, 2023), more than 70% of pharmaceutical production documentation is still paper based. “What we try to achieve now is that we can reduce the time for the recipe management from months to hours,” says Shaun Tyler, Vice President Artificial Intelligence, Software at Körber Business Area Pharma.

While pharmaceutical manufacturers have digitalized many aspects of production, recipe creation, review, and maintenance often remain heavily dependent on paper documents and manual processes. Converting those records into structured digital workflows is only the first step. The greater opportunity is making them searchable, maintainable, and continuously usable throughout the product lifecycle.

The challenge extends far beyond simply converting paper into digital files. Pharmaceutical manufacturing recipes are highly complex operational records that evolve continuously throughout a product’s lifecycle, requiring manufacturers to carefully manage changes while maintaining strict regulatory compliance.

A single pharmaceutical manufacturing site may contain hundreds or even thousands of production recipes, many spanning thousands of process steps. Updating them requires extensive manual review and deep institutional knowledge.

“You have to read through all those thousands of pages and create the very first stable recipe,” says Tyler. “This takes sometimes months…sometimes up to half a year with several people involved. Our PAS-X software helps to reduce that time drastically.”

Regulatory updates create an even greater burden. Pharmaceutical manufacturers operating under Good Manufacturing Practice (GMP) and broader GxP requirements (further Good Practice guidelines) must carefully validate and document every operational change. Even minor modifications can require extensive review across massive recipe libraries. “The difficult part is that you don’t actually know where the new regulation affects your recipes,” Tyler explains. “This might be a very long process.”

Building an AI-native platform on Microsoft Azure

As generative AI technologies matured, Körber recognized an opportunity to fundamentally rethink pharmaceutical manufacturing workflows. But bringing AI into these highly regulated production environments introduced an entirely new challenge: trust in AI.

Because manufacturing records directly influence how medicines are produced, AI-generated outputs must meet exceptionally high standards for accuracy, precision, traceability, and validation before they can be relied upon in production-critical workflows. “We can’t have the remote possibility that something is incorrect,” says Tyler.

To address this challenge, Körber worked with Microsoft in 2023 to build a next-generation AI platform for its MES and software solutions powered entirely on Microsoft Azure. “Today, all our AI products are fully built with Microsoft Azure,” says Tyler.

In pharmaceutical manufacturing, where every output must withstand GxP scrutiny, Körber built its AI platform on a governed Azure foundation. Azure OpenAI Service and Azure AI Foundry provide the generative AI capabilities and model governance, supported by additional Microsoft Azure services for scale, security, and monitoring. “Azure is deeply woven into our solution,” Tyler says. “Today we couldn’t release, and wouldn’t want to release, without it.” Azure AI Foundry also allows Körber to evaluate and govern new models as they emerge while re-validating outputs to the standards GxP requires.

“By building on Microsoft Azure, we weren’t simply digitalizing documents,” says Julia Schmitz, Head of Global Software Integration, Software, Körber Business Area Pharma. “We were creating the foundation for an intelligent manufacturing platform. Once recipe information becomes structured and trusted, customers can begin interacting with it in entirely new ways.”

The alignment with Microsoft gave Körber the scalability, performance, and cloud-native architecture required to operationalize AI across highly regulated workflows. “When you talk about getting the age of AI, it’s of course Azure AI Foundry,” says Tyler. “We are heavily dependent on good scaling and response time of the models.”

Enabling trusted AI in regulated pharmaceutical environments

At the center of Körber’s innovation is PharmaGuardrails, the company’s proprietary compliance technology designed to make generative AI usable in regulated pharmaceutical environments. Körber developed PharmaGuardrails to address one of the pharmaceutical industry’s largest barriers to AI adoption: proving that AI-generated outputs can be trusted, validated, and audited.

“No one would trust a medicine produced with a high hallucination rate,” Tyler explains. The technology wraps pharmaceutical governance, traceability, and validation controls around Azure OpenAI models, enabling Körber to validate AI outputs for regulated use cases.

According to Körber, the backbone technology enables AI output with reported precision rates near 99% for pharmaceutical tasks, according to internal testing results, while maintaining 100% numerical precision for production-critical values. That distinction matters because configuration values carry direct patient-safety consequences: if a sterilization step calls for holding material at a specific temperature and an AI system were to report the wrong figure, contaminants could survive and reach patients. Providing exact numerical values is what allows generative AI to be trusted on the shop floor.

“That means with PharmaGuardrails you can support greater confidence in AI-assisted outputs,” Tyler says.

Körber also built extensive auditability and traceability into the platform to help satisfy external regulators and customer compliance teams. “Everything needs to be traceable,” Tyler explains. “Every version we changed needs to be perfectly documented.”

Accelerating the future of pharmaceutical manufacturing

With its AI platform running on Microsoft Azure, Körber is already helping pharmaceutical manufacturers reduce manual work and accelerate MBR digitalization. In pilot implementations, Körber customers have reported recipe digitalization cycle times improving by approximately 30%, while reducing manual work by up to 40%.

The real value emerges after recipes have been digitalized. Structured manufacturing data allows engineers to search thousands of production records using natural language, understand how regulatory changes affect existing processes, and prepare updates in a fraction of the time previously required.

“Customers aren’t looking to digitalize recipes simply for the sake of digitalization,” says Schmitz. “They’re creating a living knowledge base that engineers can query, update, and continuously improve with AI assistance while maintaining the controls required for regulated manufacturing.”

The company’s AI-powered recipe management capabilities now allow engineers to interact with complex manufacturing workflows using natural language, identify regulatory impacts across recipe libraries, and support faster updates aligned to regulatory requirements that previously required months of manual effort. Underpinning these capabilities is an agent-based architecture built on Azure: a generative AI agent network reads legacy paper and PDF recipes, maps each instruction to the correct function in the MES, and structures the result into a compliant digital record ready for human review.

For Körber, the broader vision extends beyond pharmaceutical manufacturing alone.

“Our vision is not only to help manufacturers digitalize today’s processes,” says Oliver Weber, Senior Vice President, Software, Körber Business Area Pharma. “We’re building an AI-native platform that enables pharmaceutical organizations to modernize operations responsibly while maintaining the trust, quality, and compliance that patient care demands.”

By combining deep pharmaceutical expertise with Microsoft’s cloud and AI ecosystem, Körber is helping establish a new model for how highly regulated industries can adopt generative AI safely, responsibly, and at enterprise scale.

Source

Originally published by Microsoft as a Microsoft Customer Story. View the original publication.