As AI transforms clinical trials and regulatory operations, life science teams face unique challenges in adopting these technologies within GxP-regulated environments. How can companies maintain compliance while automating document processing? Where should clinical and regulatory teams begin their AI journey?
We sat down with Keith Parent, CEO of Court Square Group and RegDocs365, to explore these pressing questions. With over 30 years of experience working with pharmaceutical, biotech, and medical device companies of all sizes, Keith gives us a unique frontline look at how firms are implementing AI solutions in clinical trials and regulatory operations right now.
From automated document processing to regulatory intelligence, their discussion cuts through the hype to examine real-world applications and implementation strategies that companies can begin exploring today.
Apple Podcasts | Spotify | YouTube | Web + Others
Summary, Key Points, and Practical Takeaways
This interview has been edited for clarity and length.
Nick Capman: Are you seeing both large and small companies getting involved in AI, and what's the extent of their involvement?
Keith Parent: What I'm seeing across the board is that all the very large pharma companies recognize the potential savings from AI integration. They've put together centers of excellence and brought in high-paid talent. They've typically surveyed their entire workforce asking, 'What should we be working on and how?' This generates lots of potential projects, but they can't do everything. So they narrow it down to the top 20 or 30 initiatives that could meaningfully generate hundreds of millions in savings.
Here's the issue though: while these solutions work well for individual companies, they're point solutions—they don't get rolled out across the industry to benefit everyone. When I look at smaller companies, they're all either dipping their toe in AI or wanting to learn more about it. Some are using ChatGPT, the more adventurous ones are doing it independently, and others are bringing in consultants. We often see failures with proof of concepts because companies don't realize that the savings generated won't cover the implementation costs.
My approach to this problem is different. If I can work with a larger customer—a big pharma company or CRO—to develop a solution that makes sense at scale, we can then adapt that for smaller companies. This is particularly important because smaller companies often need these solutions more, given their limited resources.
What are your thoughts on companies using ChatGPT as a starting point?
ChatGPT is a great way to start, but you have to be very careful about using it within your firewall. You don't want your intellectual property becoming part of the public domain. Early adopters weren't as careful as they should have been about this. You also have to be aware of hallucination—people don't realize that ChatGPT might sometimes generate answers based on incomplete or inaccurate data.
I've seen an interesting shift in attitudes. In my AI and RIM working groups, many regulatory leaders were adamant: 'We would never use AI to generate documents for regulatory authorities.' Then ChatGPT emerged, and everyone started experimenting. Within a year, their tune changed to 'How fast can we generate output for our submissions?' Now the focus is on implementing guardrails and using techniques like RAG (retrieval augmented generation) to pull data from their own systems for content generation.
The key is bringing together subject matter experts from clinical or regulatory groups with AI specialists. That's how you identify real problems and develop effective solutions.
How are companies handling AI ownership when transitioning between large and small organizations?
Our approach to solutions and ownership is strategic. We don't typically work on a work-for-hire basis where clients maintain complete ownership. Instead, we develop solutions that can be incorporated into broader systems. Many AI companies I work with create what they call 'smart functions' or routines—while these might address specific problems, they're designed to be generic enough for wider application. Of course, if a company specifically pays for a customized solution, that becomes their IP.
Where should companies start with AI, and what ideas can they borrow?
Industry groups are a great starting point. I'm involved with several, including the DIA (Drug Information Association). Their AI working group—which we actually call Intelligent Automation because it better describes what we're doing—is particularly valuable. There's also the RIM reference model and the IRIS working group for regulatory matters. While some require membership, many offer webinars that anyone can access.
For those of us in the industry, it's valuable to look at what other sectors are doing and translate those learnings back to life sciences. However, if you're specifically in clinical or regulatory roles, I recommend focusing on industry-specific groups that discuss use cases directly relevant to your work.
I've noticed an important shift in attitudes during network gatherings. The barriers are coming down. People who used to say 'We'll never do that' because of quality or regulatory concerns are now asking 'How can we do that?' That's a crucial change in mindset.
What are some successful AI concepts and projects you've seen implemented?
Let me share some concrete examples we've worked on:
First, TMF Auto Classification. Think about the trial master file process: it typically takes a CRO employee 8-10 minutes to open a document, read it, determine its classification, and file it properly. By implementing natural language processing to read and understand these documents automatically, we can dramatically reduce this time while maintaining accuracy.
Second, we're looking at automating labeling information compilation. Instead of manually pulling data from multiple systems for submission labels, we're using retrieval augmented generation (RAG) to automatically gather and compile this information.
Third, there's regulatory correspondence management. Consider when you receive communications from FDA, EMA, or Health Canada about a drug product. Typically, someone reads these emails, manually identifies commitments, and enters them into a spreadsheet with due dates. We're developing systems that can automatically detect regulatory communications, archive them appropriately, and use natural language processing to identify and track commitments
What are the key do's and don'ts for companies starting with AI?
For the do's, I strongly recommend starting with low-hanging fruit. Look for processes that are highly manual and time-consuming, especially ones that have highly trained professionals doing administrative work. A clinical specialist shouldn't be spending their time pushing papers—that's a perfect opportunity for automation.
Target these simpler processes for your first AI projects. Start small and build up. Don't try to 'eat the elephant'—that is, don't tackle your biggest challenges first. A failed major project can create a ripple effect where people become resistant to future AI initiatives. Nobody wants to be tagged with 'oh, we tried that, and it didn't work.'
Let me give you a practical example from mergers and acquisitions. When companies acquire assets, they receive massive document sets. A good starting point might be using AI to identify missing documents and cross-references. Once that's working, you can expand to the next phase: automatic document classification. Can we automatically route legal documents to the legal department and financial documents to finance?
Here's another example: CROs often return trial documentation as one massive PDF file. Creating an AI tool to break this apart into properly named individual documents solves a real, painful problem that everyone recognizes. Choose projects like these—where the pain point is obvious and the benefit is clear.
Remember, success breeds success. When you're introducing new technology into a company, starting with these smaller, achievable wins helps build momentum and support for larger initiatives.
Keith's key takeaways:
Start Small, Show Success: Begin with "low-hanging fruit" - simple, manual processes that are time-consuming and don't require highly skilled personnel. Success with smaller projects builds momentum for larger initiatives.
Protect Your IP: When using AI tools like ChatGPT, ensure they're used within your firewall. Early adopters weren't always careful about protecting intellectual property - learn from their mistakes.
Scale Solutions Effectively: Large pharma companies can develop point solutions, but the real value comes from creating solutions that can be scaled down to benefit smaller companies with fewer resources.
Incorporate Human Oversight: Maintain "human-in-the-loop" verification when AI confidence levels are lower. Use machine learning to improve accuracy over time based on human input.
Focus on Time and Cost Savings: Evaluate AI projects based on concrete metrics. Ensure the savings generated from implementation exceed the costs. Failed proof-of-concepts often result from overlooking this basic arithmetic.
Leverage Industry Groups: Join organizations like DIA, RIM, and IRIS working groups to learn from others' experiences. For clinical and regulatory professionals, focus on industry-specific use cases rather than general AI applications.
Automate Administrative Tasks: Free up skilled professionals from paper-pushing. Focus AI implementation on tasks like TMF classification, regulatory correspondence management, and document processing.
Build on Success: Once a simple solution proves valuable, expand its capabilities methodically. For example, move from document identification to automatic classification and routing.
Learn from Other Industries: While focusing on life science-specific applications, look to other industries for innovative approaches that can be adapted to regulated environments.
Embrace the Culture Shift: Recognize that attitudes toward AI in regulatory environments are rapidly evolving. The question has shifted from "Can we use AI?" to "How fast can we implement it?"
The FDA Group helps life science organizations rapidly access the industry's best consultants, contractors, and candidates.
Need expert help planning and executing compliance projects or connecting with a quality and compliance professional? Get in touch with us. If we haven’t worked together yet, watch our explainer video below or head to our company introduction page to see if we could help you execute your projects now or in the future.
Our service areas:
Quality Assurance | Regulatory Affairs | Clinical Operations | Commissioning, Qualification, and Validation | Chemistry, Manufacturing, and Controls (CMC) | Pharmacovigilance | Expert Witness
Our engagement models:
Consulting Projects | Staff Augmentation | FTE Recruitment | Functional Service Program
Our podcast:
Apple | Spotify | YouTube | Web + Others
Share this post