The regulatory landscape is transforming at an unprecedented pace, and artificial intelligence is at the heart of this evolution. At the recent RAPS San Francisco Chapter event, "Compliance at Machine Speed: Unlocking the Potential of AI in Life Sciences," industry leaders gathered to share practical insights on implementing AI in regulatory affairs and quality management.
Our CEO, Michelle Wu, joined distinguished panelists Nathalie Raad, Taylor Bean, William (Al) Kentrup, and Charles R. Han for a dynamic discussion that moved beyond theoretical possibilities to address real-world applications, challenges, and strategies for AI adoption.
The Reality of AI Adoption: Progress and Challenges
Drawing from experience with over 200 companies spanning medtech to biopharma over the past five years, Michelle highlighted a critical truth: enterprise-scale AI adoption is genuinely difficult. Yet early adopters are finding success by focusing on specific, high-value use cases.
Regulatory intelligence has emerged as a primary entry point for AI in life sciences. With regulatory policies evolving constantly across global databases, AI excels at fetching insights and populating reports that support cross-functional teams and regulatory strategy development. For medical device companies, AI assists with predicate research for 510(k) submissions, while pharmaceutical firms leverage it to identify prior art and analyze assets that have received FDA approval.
From Draft to Submission: AI as a Writing Partner
While current technology cannot autonomously write complete regulatory submissions, smart regulatory professionals are finding ways to guide AI effectively. The most successful approach involves collecting relevant data and using tools like Copilot or ChatGPT to generate submissions that are "80% there", transforming rough drafts into polished documents more efficiently than traditional methods.
As panelist Charles Han noted, AI serves as a powerful communication bridge: "I speak regulatory, my leadership doesn't." By asking AI to reframe regulatory strategy for audiences with sales or engineering backgrounds, regulatory professionals can ensure their expertise translates across the organization.
Quality Management: AI Running in the Background
The most effective AI implementations in quality management are often invisible. AI-enabled quality management systems work behind the scenes to ensure compliance and consistency as human experts input data. For complaint handling, both third parties and major pharma companies deploy AI tools that classify incoming complaints, whether through verbal dialogue, written correspondence, or video, as critical, major, or minor, significantly accelerating response times.
Manufacturing investigations represent another practical application. By defining the characteristics of a quality investigation and building those parameters into AI tools, organizations can standardize and expedite the investigation process across manufacturing, clinical supplies, and development.
The Foundation: Data Quality Matters
A recurring theme throughout the discussion was the critical importance of data governance. As Al Kentrup emphasized, "It's garbage in, garbage out." Before implementing AI tools, organizations must invest time in cleaning their data sets, learning from complete response letters, problematic submissions, and regulatory feedback to create reliable training data.
This preparation extends beyond simple cleanup. It requires cross-functional collaboration that many organizations haven't traditionally practiced, bringing together regulatory, quality, IT, and digital teams in new ways.
Practical Steps for Getting Started
The panel offered concrete advice for regulatory and quality professionals beginning their AI journey:
Start small and build trust. Begin with straightforward tasks like gap analyses or flowcharts. The work you might assign to a new hire. This allows teams to understand how AI responds and make corrections before tackling more complex applications.
Leverage existing tools. Rather than building everything from scratch, look to vendors with established expertise in translation, data management, regulatory databases, or intelligence. These platforms offer quick wins through natural language searching and filtering capabilities that were previously cumbersome.
Test for consistency. When adopting AI, ask the same question multiple times within a reasonable timeframe. If you receive three different answers, proceed cautiously. Always request sources and verify them independently.
Choose your tools strategically. Michelle shared her perspective on the current AI landscape: ChatGPT excels at writing, Claude (Anthropic) demonstrates superior reasoning capabilities, and Gemini leads in video editing, voice, and image processing. For enterprise environments, Microsoft Copilot offers the advantage of potentially passing security checks more easily, though it essentially functions as an LLM wrapper with model flexibility.
Navigating the Security Landscape
Perhaps the most pressing concern raised during the event centered on data security—a critical consideration when feeding proprietary submission data into AI tools.
Michelle provided clear guidance for different use cases:
For personal exploration: Use your personal browser, computer, and Wi-Fi to experiment with various AI tools. Be aware that some platforms, including ChatGPT's new browser, use all queries for training purposes, including competitive intelligence questions that could inadvertently inform competitors.
For professional use: Take three essential steps before adopting any AI tool:
- Review privacy policies thoroughly
- Understand data usage practices
- Confirm data deletion capabilities
For consultants and contractors handling client data, these considerations become even more critical given existing confidentiality agreements and MSAs.
Looking Ahead: The Convergence of Regulatory and AI Tools
An intriguing possibility emerged from the discussion: should companies adopt the same AI tools that regulators use? With FDA utilizing specific platforms, some organizations are considering matching those tools to ensure their submissions are "AI ready", effectively speaking the same language as reviewers and anticipating what regulators will see when they apply their own AI analysis.
The Path Forward
As Michelle noted, foundational AI models are rapidly becoming commoditized as tech leaders acknowledge they've nearly exhausted available internet information. The competitive advantage will increasingly come from implementation strategy, data quality, and organizational readiness rather than the choice of underlying model.
For regulatory and quality leaders, who are universally "overbought, overworked, and understaffed", AI offers genuine relief. Whether it's quickly sifting through international data to find precedent, reformulating complex regulatory arguments for different audiences, or simply transforming lengthy, comprehensive essays into concise, smart executive summaries, AI amplifies the thoughtful expertise that regulatory professionals bring to their work.
The message from the RAPS event was clear: AI in life sciences compliance isn't coming. It's here. The question is no longer whether to adopt AI, but how to do so thoughtfully, securely, and effectively. Those who start small, build trust gradually, and maintain rigorous data governance will be best positioned to unlock AI's potential while navigating its risks.