- David Jones
- Jan 14
- 3 min read
Insights from Paulius Ojeras at COG Bay Area – Redefining Efficiency, Quality, and Oversight with AI Teammates
Rethinking Outsourcing in the Age of AI
In an era defined by digital acceleration and increasing trial complexity, clinical operations professionals are under pressure to deliver faster, higher-quality trials without expanding resources. At The COG Review: Building Better Clinical Studies, Paulius Ojeras - an industry veteran and Head of Clinical Operations at Perceive Biotherapeutics - shared an actionable, firsthand perspective on the real impact of AI in trial operations. The session, hosted by David Jones, drilled down into not just “if” but “how” organizations can adopt AI agents thoughtfully, maximizing operational efficiency without sacrificing oversight or patient safety.
The Five Levels of AI: Where Are We Now?
Paulius Ojeras introduced a nuanced framework for understanding AI’s current and future capabilities in clinical research. AI tools today range from basic automation (simply triggering and verifying routine steps) to advanced agentic systems, capable of planning and delivering outcomes autonomously with minimal human intervention. Most biotechs, he noted, are currently operating between levels one and three – using AI for automation and smart assistance, always with a human in the loop.
The real shift is beginning at levels four and five, with “agentic AI” acting as digital teammates. These agents can coordinate, execute, and oversee clinical operations much as a CRO (Clinical Research Organization) team would, but with unprecedented scalability, reliability, and resistance to human error. The big question is not when, but how, AI teammates will become fully embedded in day-to-day trial delivery.
Real-World Application: AI Teammates in TMF Management
To ground the discussion, Paulius Ojeras shared a detailed case study from Perceive Biotherapeutics’ own transformation. Facing significant quality gaps and resource backlogs in their Trial Master File (TMF) management, the team decided to pilot an AI-managed approach.
The results were striking. When comparing CRO-processed TMFs against those handled by AI teammates, 25% of “final” documents flagged by the AI had quality issues - ranging from missing dates to improper file naming. These findings were not only consistent across organizations large and small but also underscored the prevalence of error in legacy setups. Beyond error detection, the AI-based system could automatically generate expected document lists, create placeholders, monitor real-time updates, and even flag document issues for immediate correction - delivering “quality by design” with live audit readiness.
This experiment demonstrated that AI agents can do more than support staff - they can redefine the process, reduce backlog, and secure compliance at a pace unattainable through human effort alone.
Vendor Selection and AI Literacy: Guardrails for Success
The transition to AI-empowered operations demands both caution and curiosity. Paulius Ojeras emphasized that not all “AI solutions” are created equal. Professionals must vet vendors for clinical research pedigree, proven deployments (not just pilots or demos), financial viability, and hands-on support. Critical questions include: Is the AI model adaptable to your workflows? How is consistency of output ensured? Can you pilot the solution before full adoption?
Importantly, he highlights that successful AI adoption is not just technical – it’s cultural. Teams must invest in AI literacy and internal prompt engineering skills. As one audience member commented, the best tools delivered poor outcomes if staff aren’t trained to interact with them effectively. “Garbage in, garbage out” applies more than ever, especially as AI agents move into areas like medical monitoring or data mining.
Rethink, Don’t Just Automate: Shaping Processes for AI
Perhaps the session’s most practical takeaway: don’t just drop AI into existing steps – redesign the workflow entirely. Integrating agentic AI means rethinking processes so they’re self-sustaining and scalable, with human expertise reserved for setup, oversight, and complex decision-making.
In highly regulated environments, consistency and accountability are paramount. It’s essential to clarify who’s responsible for decisions, how outputs are validated, and what manual checks remain necessary. With the right guardrails, AI can amplify – not complicate – organizational capability, freeing staff to focus on patient-centred goals.
Conclusion: The Future Is Collaborative
As the adoption curve steepens, Paulius Ojeras urges organizations to move thoughtfully, balancing innovation with robust oversight. In his words: “True success is not how fast we incorporate AI, but how smart, thoughtful, and disciplined we are in using these solutions.” Clinical trial professionals ready to embrace this collaboration between human and artificial intelligence will lead the way to smarter, more efficient, and ultimately, more patient-centered research.
Comments