Agentic RTSM and the Push to Compress Clinical Trial Startup: Q&A with Robert Hummel, Suvoda

Clinical trial startup has long been one of the most resource-intensive phases of study execution, with technology configuration cycles consuming time and attention that sponsors could otherwise direct toward site readiness and patient recruitment.

To explore this further, Applied Clinical Trials spoke with Robert Hummel, chief operating officer at Suvoda, about where the biggest operational bottlenecks persist, how agentic artificial intelligence (AI) is changing the build and deployment process for randomization and trial supply management (RTSM), and what safeguards are needed to accelerate timelines without compromising quality or compliance.

1. What is driving the industry’s push to reduce clinical trial startup timelines, and where are the biggest operational bottlenecks today?

Hummel: Startup is still one of the hardest windows to compress. Getting from protocol finalization to system go live can take weeks, even as pressure grows to manage rising protocol complexity and ultimately bringing therapies to patients faster. Sponsors and CROs are being asked to tackle more complexity on the same or shorter timelines.

My focus is on identifying the bottlenecks, especially those that are repeatedly an issue:

• Extended, up-front requirements and specification work for study technology.
  Traditional randomization and trial supply management (RTSM) technology builds historically have begun with a high burden of extensive documentation. Teams turn the protocol into detailed requirements, iterate the language, and secure sign off before the build can start. This consumes a disproportionate amount of attention at the exact moment they are also managing site selection, regulatory submissions, and patient strategy. It also creates the potential for misalignment on how the system will ultimately work, which can lead to rework and further delays.
• Repeated manual configuration, customization, and testing.
  After requirements are approved, configuration, study specific customizations, and testing have historically relied on highly skilled, manual, time-intensive work that can take a month or more, repeated study after study.
• Mid-study change orders and protocol amendments.
  When protocols shift, change orders can feel like a second startup. Requirements must be updated, systems reconfigured, and tests rerun. Each cycle can delay moving forward with study changes, slow the study, or add risk.

The result is a startup window that is both a persistent drag on timelines and one that can impact a team's ability to get the study up and running while consuming significant effort from sponsors. At Suvoda, we are providing a solution with agentic RTSM.

2. How could AI and automation change the way RTSM and other trial systems are configured, tested, and deployed?

Hummel: AI should take on the repeatable operational work, not replace human judgment.Agentic RTSM uses specialized AI agents to handle configuration, customization, and testing under human oversight, instead of completing each task manually.

Agentic RTSM is changing the approach:

• From long-term requirements documents to an intuitive, show-and-tell build.
  Sponsors have been asked to complete long specification documents and then wait to see how that translates at UAT, which can be a month or more later.With agentic RTSM, we can move to a more visual and collaborative process. Sponsors can react to what they see, based on their initial inputs which form the basis of the study build, and then refine until the system matches what they need. The visual specification experience makes it much easier to reach the right design without cycles of paperwork and rework.
• Configuration and customization as agent tasks.
  Once the study design is clear, a team of AI agents can get to work. At Suvoda, our specialized XD Agents handle study-specific configuration and customization tasks. These agents draw on proven configuration patterns from thousands of prior trials, and our domain experts stay in the lead to guide and validate how those patterns are used for each new study.
• Automated testing and validation with full oversight.
  A Test XD Agent then generates and executes test cases and records evidence, so human experts can focus on edge cases and complex logic rather than building every test script. Every agent action is logged and auditable, and nothing moves forward without human review and sign off.

All of this is enabled by the Suvoda Platform. It harnesses two important innovations: a patented Virtual Partition architecture separates customizations from core code, and low-code/no-code tools which make customizations more straightforward and standardized. Agents are constrained to the study-specific configuration layer and cannot change validated core code. That allows us to safely compress build and deployment timelines in RTSM and, over time, in eCOA and financial services as well, while maintaining the level of control and traceability that sponsors and regulators demand.

3. What impact could faster study startup have on site readiness, patient enrollment, and overall trial execution?

Hummel: RTSM speed alone does not guarantee earlier first patient in. Startup is influenced by many factors. When technology is no longer one of the slowest and most demanding workstreams, it can change how sponsors and CROs experience that part of the trial because every week of delay in startup is a week patients are still waiting.

The impact is meaningful and includes:

• More attention available for site readiness.
  If RTSM build moves from a heavily manual effort to a roughly two-week, agent-assisted process, then teams can get back weeks of focused attention. They can reinvest that time in site selection, training, and operational planning, which are the activities that truly determine how ready sites are on day one.
• Better preparation for enrollment and patient experience.
  When systems are ready earlier and require less hands-on management during a build, sponsors can invest more time in patient recruitment strategy, feasibility work, and communication with investigators and participants, rather than spending it on configuration cycles and document review.
• Smoother execution and responsiveness mid-study.
  The same agentic patterns that accelerate initial build also shorten the cycle for change orders and protocol amendments. That can compress timelines for study changes and help teams manage protocol amendments without delaying study execution.

Faster and less burdensome technology startup can free up capacity in exactly the areas where human judgment makes the most difference.

4. As AI becomes more embedded in clinical trial operations, what safeguards are needed to balance speed, quality, and oversight?

Hummel: Speed and quality are not tradeoffs. Responsible acceleration in a regulated environment depends on several safeguards which are nonnegotiable. At Suvoda, our approach is:

• Human in the loop design and sign off.
  Our AI agents are intentionally designed as assistants to domain experts, not replacements. People define rules, review agent outputs, and decide when a build is ready to move to UAT or production.
• Auditability and compliance by design.
  Every step of an agentic RTSM build needs to be logged and auditable. Compliance is built into the platform, rather than being added as a separate layer.
• Data protection and zero retention practices.
  We run a closed, enterprise environment. Customer trial data is never used to train public or third-party models. Agents and our assistant Sofia follow a zero-retention approach, keeping data only long enough to fulfill a given request. Access is role based at the study level so both users and AI agents see only what they are authorized to see.
• Domain-specific knowledge and validation frameworks.
  A proprietary knowledge base is informed by thousands of complex trials. A “master-apprentice” agentic architecture allows domain experts to guide and shape agent behavior, and that supports the validation and audit trails.

The result is accelerating the workflow while maintaining full traceability, control, and alignment with patient safety and regulatory expectations.

5. Looking ahead, how do you see agentic AI and intelligent automation reshaping the broader clinical trial technology stack?

Hummel: We are moving to a world where AI organizes how the trial technology stack is built, operated, and improved.

Suvoda is leading trends with meaningful solutions:

• Embedded conversational intelligence.
  Sofia, our AI assistant, gives sponsors and sites instant, conversational access to data, currently in RTSM and expanding to eCOA and beyond. Sofia is beginning as a question-and-answer assistant and is evolving toward a more robust agentic role, with the ability to help teams take action in the system.
• Agentic AI will be implemented across multiple products.
  Our XD Agents compress build, testing, and amendments. Our roadmap extends to eCOA and financial services, so everything from questionnaire flows to participant payments/travel can benefit from automation and consistency.
• A move toward unified, action-driven platforms that streamline trial operations.
  AI assistants and agents are even more powerful if implemented on unified software platforms, like the Suvoda Platform, which is designed to synchronize activities and data in real time with a single, well-structured data layer. The platform foundation allows AI agents and assistants to operate safely across trial activities, while maintaining regulatory compliance, data integrity, and full auditability.

The long-term direction is a technical stack where build, execution, and insights are all AI enabled, and where human expertise, patient safety, and regulatory expectations remain at the center. Agentic RTSM is an early example of how AI will fundamentally reshape clinical trial operations over the next decade.