Quality Leaders Warn Against Digitizing Broken Processes

By Allison Proffitt 

August 18, 2026 | BOSTON — At last week's Bioprocessing Summit, Gilead Sciences' Senior Vice President of Global Quality Anthony Mire-Sluis opened the Wednesday plenary session with a blunt warning to an industry rushing toward AI: digitizing a bad process just gets you a bad process, faster and more expensive to fix. 

Mire-Sluis argued that many pharmaceutical companies are chasing AI and digital transformation without first cleaning up the underlying workflows — and that doing so backfires. His central example: a 55-step paper approval process that gets turned into 55 digital approvals accomplishes nothing, since typing a password takes as long as signing a form. Worse, he said, digitizing a complex paper batch record often strips away flexibility operators relied on, turning tasks that could be done in parallel into rigid, sequential steps that make life harder on the plant floor, not easier. 

He laid out three imperatives for getting digital and AI investments right: fix the process first, get the data right, and manage the change.  

Fix the process first. Once a system is validated in a GMP environment, changing it means revalidation and re-testing. It’s much wiser to simplify before digitizing than after. He described a five-step approach: define the desired outcome, map the process with the people who actually run it in person (he was emphatic that this doesn't work over Zoom), identify friction points, redesign the workflow, and build in room for post-launch adjustments. 

Get the data right. AI and digital systems are only as good as what feeds them, he said, citing an internal Gilead mock recall in which the same bottle was labeled three different ways across three sites, turning a routine trace into a multi-system scavenger hunt. He called for company-wide data standardization, clear data ownership, and formal governance — unglamorous work, he acknowledged, but foundational. 

Manage the change. Without a clear explanation of why a digital or AI initiative matters, employees tend to assume it's a headcount-reduction exercise, Mire-Sluis said, which breeds resistance. He pointed to Kotter's eight-step change model as a personal touchstone and described Gilead's AI office hours that invite any employee to ask questions, ask for help, or share a win in an internal "build and brag" program, where staff share self-built AI agents, as ways to build enthusiasm rather than fear. 

He offered a concrete example of payoff: an AI tool that scans roughly 30,000 historical deviations in seconds and ranks them by relevance — work that used to consume enormous staff time on manual keyword searches. 

Fireside Chat: Pfizer, GSK, Gilead On Standardization's Hard Lessons 

Mire-Sluis was then joined by Susan Hynes, Global Head of Quality at GSK, and Lynn Bottone, Senior Vice President of Quality Operations and EHS at Pfizer, for a discussion about their own digital transformation efforts. 

Bottone, a 36-year Pfizer veteran, described an early deviation-management rollout across the company's 37 internal sites that went badly — precisely because site staff weren't brought into the design process and a flawed workflow was forced into the new system. She said Pfizer has since paused and restarted its approach, now called "Revolution," standardizing global procedures with local work instructions layered on top, tackling the quality management system in full rather than piecemeal. 

Hynes described a similar reckoning at GSK, which was merging quality systems from what had effectively been two separate companies (vaccines and medicines) with different cultures and terminology. GSK's original "three S's" approach — standardize, simplify, seamlessly integrate — leaned too heavily on top-down standardization, she said. The company has since rebalanced toward simplification, using AI to help identify which process steps are actually critical to compliance and which can be eliminated. She cited a same-day example: a team had just removed 58 non-GMP-critical documentation steps after concluding AI made them unnecessary. 

Both executives emphasized workforce messaging. Pfizer has rolled out tiered AI training (AI 101 and AI 201) and floated the framing “agents work, humans judge” to describe how roles will shift rather than disappear. Bottone noted that a QP's job — verifying dozens of naming conventions for the same raw material, for instance — is exactly the kind of drudgery AI can absorb, freeing staff for higher-judgment work. Hynes echoed the point: the goal, she said, is collapsing what might be three systems and dozens of screens into a single screen for the person releasing a batch. 

Where do these solutions come from? Hynes said GSK initially over-invested in internal customization and has since shifted to a mixed model — keeping quality assurance work customized in-house, but leaning on external vendor partnerships for quality control, where GSK is only beginning to focus and can't move fast enough alone. 

Cultivating Buy-In 

The speakers had already reflected their own challenges getting buy-in from their teams when an audience member asked how to build grassroots AI adoption at smaller, less-resourced organizations. Hynes advised starting with a clear, simple problem statement, since teams often disagree for hours about what problem they're even solving. She also described setting an intentionally ambitious, non-punitive target at GSK — cutting 30% of waste in a year — without prescribing how, arguing that incremental goals produce only incremental (and often more complex) change, while a bold, unmeasured ambition gave teams permission to rethink processes from scratch. 

All three leaders acknowledged that hardest parts of AI adoption in a regulated GMP environment aren't the algorithms — they're process discipline, data standardization, and convincing a workforce that automation is additive rather than a threat to their jobs. 

They recommended frequent recognition: informal shout-outs, town hall demos, and programs like Gilead's "build and brag" that turn early AI adopters into visible internal champions who bring colleagues along.