The Design Decision Problem
Designing a clinical trial is a series of interdependent decisions made under significant uncertainty. How many visits should the schedule include? Which procedures at which visits? How do these choices affect patient burden, site workload, and study cost? What does the literature suggest? What have similar studies done?
These questions require expertise that is distributed across clinical scientists, medical writers, data scientists, and operational specialists. The knowledge exists - but accessing it in real time, in the context of a specific study design decision, has historically been impossible. Teams rely on institutional memory, manual literature searches, and ad hoc consultation with colleagues. The result is design decisions that are less informed than they could be, and a process that takes far longer than it should.
"Franklin is not a chatbot. It is a purpose-built intelligence system grounded in clinical study data, standards, and decades of institutional knowledge."
What Franklin Is - and Isn't
Franklin is Trials.ai's AI intelligence layer embedded within Smart Designer. It is not a general-purpose large language model. Franklin is a purpose-built system grounded in three foundational data sources that give its responses clinical validity and specificity:
- The Clinical Studies Ontology: 500K+ classes connecting clinical study concepts across standards and terminologies. Franklin uses this to understand the semantic relationships between study elements - what an endpoint means in the context of a therapy area, what procedures are typically associated with it, and what constraints apply.
- The Benchmark Database: Structured data from thousands of historic clinical studies. Franklin uses this to provide evidence-based comparisons - showing how the current design compares against similar studies in terms of schedule complexity, visit count, assessment burden, and operational characteristics.
- The Patient Burden Index: A validated scoring model that quantifies the burden imposed on patients by study procedures. Franklin uses this to flag high-burden design choices and suggest alternatives grounded in evidence.
This grounding is what distinguishes Franklin from a generic AI assistant. Franklin cannot hallucinate endpoint recommendations - they are drawn from a validated ontology. Franklin cannot invent benchmark comparisons - they are computed from structured study data. Every response is traceable to a data source.
Three Core Capabilities
Guided Exploration
Ask natural language questions about your study design and get contextually relevant answers grounded in your data. "How does our visit count compare to similar oncology studies?" "What endpoints are typically included in Phase II studies for this indication?" "What is the burden impact of adding this assessment?" Franklin draws on ontology, benchmarks, and study context to provide specific, evidence-based responses - not generic suggestions.
Tradeoff Analysis
Model design scenarios in real time. Franklin visualizes the cost, burden, and schedule implications of design alternatives, enabling teams to make better decisions faster. Change an endpoint, add a visit, remove an assessment - and see the downstream impacts immediately, alongside benchmark comparisons that show how each alternative performs relative to similar historic studies.
Structured Recommendations
Receive specific, actionable suggestions to optimize your protocol. Franklin identifies high-burden procedures, schedule inefficiencies, and design patterns associated with protocol amendments or operational failures - and recommends specific changes with supporting evidence. Recommendations are structured, prioritized, and linked to the ontology concepts they reference.
Integration with Smart Designer
Franklin is embedded directly within Smart Designer - not a separate tool that sits outside the workflow. As clinical scientists build their study design, Franklin provides real-time guidance at each step, in the context of the specific choices being made.
When a scientist is building the Schedule of Activities, Franklin surfaces relevant benchmark comparisons. When they're finalizing eligibility criteria, Franklin flags potential recruitment challenges based on historic enrollment data. When they're reviewing the completed design, Franklin provides a structured assessment with specific optimization recommendations - before the protocol reaches IND submission.
This in-context integration is critical. The value of AI guidance is highest at the moment a decision is being made, not after the fact. Franklin is designed to be present at every decision point in the study design process.
What This Means for Clinical Teams
Franklin changes the nature of clinical study design from an expertise-intensive, time-consuming process to an intelligence-augmented workflow. Senior clinical scientists can focus their expertise on the decisions that truly require it - the scientific judgment calls that no algorithm can make - rather than on data gathering, benchmark searching, and document reconciliation.
Junior scientists and newer team members have access to evidence-based guidance that accelerates their development and reduces the risk of design errors that would otherwise require amendment. Organizations with high turnover or distributed teams benefit from consistent, institutionalized knowledge that does not walk out the door when experienced staff move on.
The goal is not to replace clinical judgment. It is to ensure that every design decision is informed by the best available evidence - and that the expertise embedded in historic studies, validated scoring models, and the Clinical Studies Ontology is accessible to every team, on every study, without requiring a specialist consultation for every question.