The ARG Approach to Generative AI in Clinical Research
April 9, 2025At Atlantic Research Group (ARG), we are excited to explore the possibilities of generative AI (genAI) within clinical research. We recognize AI's transformative potential in this space while remaining mindful of data security and hallucinations (situations where genAI platforms create fictional information).
To address these concerns, ARG has established a robust AI usage policy. We prohibit the use of confidential information in free AI tools and have adopted Google Gemini through our secure Google Workspace, allowing safe experimentation. The data from the chats in this environment are not used to train the Gemini models. Google Gemini's AI privacy policy can be found here.
Additionally, we have formed a genAI users group made up of individuals from every department at ARG. These individuals will serve as advocates for the technology while providing constructive feedback. They will help us properly experiment within our Gemini Google Workspace environment and learn to incorporate genAI into our current processes while simultaneously creating new processes from expanded capabilities. This group will also serve as liaisons to others who are learning the technology to help them utilize it effectively without making the same errors we all do when starting with generative AI. While the technology continues to evolve, members of this group will help their colleagues understand what generative AI excels in and what it is not yet equipped to do.
Currently, the Gemini chat app serves as a powerful brainstorming partner and writing assistant, helping us refine ideas and improve communications. In the very near future, it's going to help the ARG team write code and scripts to create graphs and reports and other high-quality outputs using natural language.
Another tool we are exploring is Google NotebookLM Plus, which is already integrated into our workspace. NotebookLM makes it possible to produce summaries, reveal insights, and ask questions about specific documents, websites, and even audio/video files. After uploading these sources, users can leverage NotebookLM as a research assistant to summarize, find details, and answer questions based solely on the provided content, greatly reducing the possibility of hallucinations. While excellent for quickly grasping and reviewing key points in large documents, these tools do not replace the human element needed for thorough reading and analysis.
Beyond our internal exploration of genAI, we are actively engaging with sponsors and vendors to understand their initiatives with this technology. For instance, one sponsor is utilizing an internal AI tool to analyze trip reports from our Veeva Vault CTMS instance. Similarly, Medrio, a long-standing EDC vendor partner, recently shared their genAI integration efforts in an informative webinar, highlighting a "crawl-walk-run" implementation philosophy that resonates with our own approach.
At ARG, we see our top-down, bottom-up approach to genAI as an opportunity to enhance our operations. From streamlining processes to identifying new research sites to improving quality control, our goal is to empower our talented team, enabling them to achieve more and take on new challenges. For us, genAI isn’t just about efficiency; it’s about expanding our capabilities and ultimately delivering a better product.
Follow our site for regular updates on how ARG is leveraging generative AI to improve clinical research. We'll share insights, best practices, and real-world examples of how we're increasing quality and improving outcomes.
