September WLW With Ulrich Köllisch: AI, CAPA and Data Governance

Why do the same data integrity problems keep appearing in inspections? That question ran through September’s Warning Letter Wednesday. On September 30, 2026, KeyPedia’s Zephaniah Odidika joined Ulrich Köllisch, Founder & Principal Consultant at umk consulting GmbH, to discuss the 2026 PDA/FDA Joint Regulatory Conference in Washington, D.C., and the PDA data governance workshop that followed.

Ulrich also leads PDA’s data integrity and data governance interest group. He brought examples from the conference and workshop to a discussion of what quality teams can do differently, from investigating missing records to reviewing a contractor’s data.

Watch the September WLW Replay

Watch the full conversation below.

AI Oversight Starts With Three Written Answers

The speakers discussed AI as a way to help investigators work through large volumes of documentation. Zephaniah’s takeaway from the conference was that investigators still collect documents, evaluate the evidence and draw their own conclusions.

Ulrich said human accountability came up repeatedly at the conference. For companies using AI, the speakers recommended writing down answers to three questions:

• What does the tool do?
• What data does it act on?
• What decisions does it inform?

Writing those answers down makes intended use, validation and human review easier to discuss. Ulrich also raised questions about confidentiality and how much detail had been shared about the controls behind the tools.

A CAPA Should Reach Beyond the Finding

Ulrich returned to a common investigation weakness: correcting the visible issue without understanding why it occurred.

A missing raw-data record, for example, may be a symptom of an outdated system, a poorly designed process or a culture that discourages reporting problems. Replacing the record or retraining an operator can leave those conditions in place.

Ulrich urged teams to widen the investigation. If one system is outdated, where else are similar systems used? If one batch has a data integrity issue, could the same failure affect other batches, products or sites?

A question for your next quality meeting: If this finding is a symptom, where else could the same cause be hiding?

Data Governance Includes People and Culture

Drawing on the PDA workshop and a technical report under development at the time of the webinar, Ulrich described four domains of data governance:

• Organization and structure.
• People and culture.
• Processes and procedures.
• Systems and technology.

He noted that organizations often concentrate on procedures and software while giving less attention to ownership, accountability and quality culture. A technology upgrade may help, but teams also need to know which data they own, what constitutes the original record and how concerns move between departments.

He suggested talking to operators, using Gemba walks and bringing quality teams and business process owners into the same conversation. Often, asking who owns a data set reveals that the responsibility is less clear than expected.

Small Devices Can Expose Large Gaps

One case discussion focused on records from small or movable instruments, including particle counters and filter integrity testers. The concerns included undocumented testing, broken traceability and unofficial records outside controlled processes.

Ulrich highlighted these devices as useful places to look during data integrity audits. The questions are straightforward: Where is the original data? Can a reviewer reconstruct what happened? Are all tests captured, including failures and retests?

The fixes discussed included controlled documentation, clearer ownership and automated transfer to a central data repository, so records remain available for review.

Follow the Data Across Your Suppliers

The conversation ended with third-party oversight. Ulrich encouraged teams to consider both the materials and the data exchanged with contract organizations and service providers.

Receiving a final result does not necessarily provide enough visibility into how it was produced. The underlying data, method validation, review responsibilities and access arrangements also deserve attention in contracts and audits.

Ulrich was clear on accountability: a company can outsource the work, but it still needs to understand and stand behind the data supporting its product.

Presentation Slides and Speakers

View or download the September WLW presentation slides.

Featured speaker: Ulrich Köllisch, Founder & Principal Consultant, umk consulting GmbH.

Host: Zephaniah Odidika, KeyPedia.

Explore more KeyPedia webinars and quality insights.

This recap reflects the September 30, 2026 webinar and the speakers’ conference and workshop observations. References to developing initiatives reflect their status as discussed during that session.