A recap of the LinkedIn Live Webinar, Strategy, Governance & Validation: What Life Sciences Needs to Win in AI
Bryan Ennis
Co-Founder & Chief Quality Officer
Sware
Martin Heitman
Secretary of ISPE GAMP Global SIG Software Automation & AI
Managing Director of MH Consulting & Advisory
AI pilots are easy to celebrate, but scaling them in a GxP environment is a different challenge entirely. That tension shaped a recent LinkedIn Live conversation we had, where we worked through the tough realities: governing AI without burying it in documentation, validating systems with endless test paths, and navigating rules that keep shifting.
Here are the takeaways that still guide how we think about it.
Many organizations fought hard to get a single AI use case through the GxP gate—and succeeded, but only after hitting wall after wall. The real lesson wasn't that AI doesn't work. It was that ad hoc, project-by-project execution doesn't scale.
Resources matter, but structure matters more. Even well-funded AI efforts can stall without a clear approach and proven guardrails. The organizations making real progress do not treat AI as a single use case to check off. They manage it as a portfolio of investments tied to the future state of the business, which is what turns spending into lasting value. If you're evaluating AI as a tool to install rather than a capability to build, you're already behind.
Organizations often start with the tool instead of the transformation. Saying, “We’ll roll out Copilot and get people to use it,” may sound like progress, but it’s really just a deployment plan—not a strategy. AI adoption only creates lasting value when it starts with a clear view of how the business process itself needs to change. The transformation has to come first. The tools should follow.
The same principle applies to governance. It should not be a polished framework disconnected from real work, but a living system that evolves alongside actual use cases. Governance sets the boundaries, use cases test them, and experience and data help refine and expand them over time.
Regulators are making progress—but guidance will always lag technology. The smartest move is to anchor on common principles and a shared understanding of "what good looks like," then apply them in a reasoned, evidence-based way.
The joint EMA/FDA AI best practices reinforce this: governance is central, flexibility is real, but it isn't free exploration. Document your intended approach and controls.
On Annex XXII: The initial draft drew hard red lines—no probabilistic models, no dynamic systems, no LLMs—which sparked a productive industry dialogue. Regulators are now reconsidering some of those positions, particularly as they learn about control techniques like guardrails. Engage in that dialogue and stay current. The frameworks are still taking shape.
AI is forcing life sciences organizations to rethink validation and take data far more seriously, and that was one of the first things we kept circling back to. Traditional paper-first, reactive approaches are not built for systems that can be tested in countless ways. Teams now have to make risk-based decisions about what actually matters to verify, from performance to bias to cybersecurity, and that requires a more mature quality culture.
At the same time, AI exposes every weak spot in data quality, infrastructure, process stability, and governance. That is why we believe validation strategy can no longer stand apart from data strategy. Without strong, well-governed data, it becomes nearly impossible to build a sustainable verification approach or scale AI with confidence.
Governance, systems, and data all converge on one truth: this transformation succeeds or fails on people. Your people are already busy. Piling on tools and training without lifting existing burdens breeds resistance. Effective change management means:
Granting space and time to build genuine AI literacy.
Building confidence in quality assurance teams who will face the inspector.
Using common language to reduce confusion.
Engaging both champions and skeptics—you need both to get implementation right.
When people are empowered to solve their own pain points, adoption becomes something they drive, not something done to them.
AI in life sciences is no longer a question of if but how well. The organizations moving to scaled deployment treat governance, validation, data, and people as one integrated challenge.
Lead with strategy, then choose tools.
Grow governance and use cases together through evidence-based iteration.
Anchor on principles, not checklists that will always lag the technology.
Modernize validation from paper-first to data-centric.
Invest in data foundations as the lever for every AI decision.
Empower your people with space, confidence, and a common language.
The organizations building strong foundations now will be the ones defining what good looks like for the rest of the industry. In our view, that starts with an honest look at where those foundations still need work.