How are you and Formation Bio using AI to accelerate clinical development?
We are building many AI tools and systems across the drug development lifecycle to help improve efficiency and outcomes, and I’ll share one example here. One example of an AI system we’ve developed is Muse, which was built to accelerate clinical trial development by analyzing scientific literature, real-world evidence, and disease insights. It can:
- Compile deep research on disease areas, patient populations, and the competitive landscape
- Identify ideal patient profiles and recruitment strategies, with a focus on a diverse breadth of patient populations and real-world representation
- Auto-generate high-quality recruitment materials and pre-screening questionnaires, customized for different audiences, channels, and languages informed by regulatory guidelines
Tasks that once took large cross-functional teams months to complete can now be done in minutes—without sacrificing quality or strategic depth. Muse is one of the first AI applications of its kind in the pharmaceutical industry and shows how well-designed AI tools can drive real, measurable impact in clinical development.
Tell us about the work that you're leading at Formation Bio.
We're leading the charge in reimagining drug development from the ground up using AI. We're an AI-native company, meaning we’re not just layering AI onto existing processes—we’re fundamentally rebuilding how drug development is done. That means asking: what would drug development look like if it were designed today, with AI at its core?
We’re applying AI to every stage of the process—whether it’s patient recruitment, protocol design, or data analysis. These are traditionally resource-intensive tasks that can take full teams and long timelines. Our goal is to dramatically streamline that work, making it possible for smaller, more nimble teams to generate insights faster and more accurately than ever before. It’s about replacing manual, fragmented workflows with AI-augmented systems that drive speed, quality, and ultimately better outcomes for patients.
What gaps in clinical development did you experience in your previous roles where AI could have been helpful?
One of the biggest limitations was the time and manual effort required to gather, organize, and interpret data to inform clinical strategy. You often have to rely on what's available through PubMed, key opinion leaders, and a patchwork of structured data sources. It's a slow, incomplete, and resource-heavy process. With AI, there's an opportunity to access and synthesize a much broader set of data—both structured and unstructured—at scale. AI can act as a continuous research assistant, rapidly aggregating and contextualizing global scientific and medical knowledge so you can make better-informed decisions, faster.
Another critical gap is in understanding the patient journey. AI can help create a far more detailed and dynamic view of how diseases progress in the real world, what treatments patients are actually receiving (on- and off-label), and which patient subgroups are most likely to benefit from specific interventions. This is incredibly valuable for designing better trials, selecting the right sites, and understanding how to position a therapy in an increasingly crowded market.
There are also major opportunities beyond trial execution—in areas like medical affairs, HEOR, and competitive strategy—where the precision, speed, and scalability of AI can create real value. Ultimately, AI has the potential not just to compress timelines, but to fundamentally enhance the strategic quality of clinical development.
How can we think about using AI to streamline clinical development?
Think of AI as a sophisticated thought partner in clinical development. It can help surface insights faster, improve data synthesis, and streamline decision-making across the development lifecycle. When applied thoughtfully, AI has the potential to reduce timelines, increase precision, and support better strategic decisions—especially in areas like protocol design, patient population analysis, and evidence generation.
AI can play a critical role across multiple points of development. It enables teams to move faster and with more confidence by automating research, uncovering patterns in large datasets, and generating new perspectives that would otherwise require significant time and manual effort. As clinical programs become more complex, AI offers a way to bring greater speed, scale, and consistency to decision-making without sacrificing scientific rigor.
Regulators, including the FDA, are actively engaging with the role of AI in drug development. CMOs have an important opportunity to lead this conversation: by clearly communicating how AI is used, how data is generated and interpreted, and how underlying models are designed. AI does not have to be treated as a black box; it can be part of a rigorous, collaborative approach with regulators.
With cost in mind, how should CMOs think about where to invest in AI?
Start by identifying the specific problems they’re trying to solve. AI is most valuable when it drives clear, measurable efficiency—doing the same work faster, better, and often at lower cost. For resource-constrained biotechs, that efficiency can be a strategic advantage. If a tool can deliver equal or better outcomes for less investment, it’s worth serious consideration.
AI also helps level the playing field. Smaller companies now have access to tools and capabilities that were once limited to large pharma. And because smaller biotechs tend to be more agile, they’re often in a better position to implement AI quickly and effectively—without the drag of legacy systems or bureaucratic inertia. That combination of cost efficiency and organizational flexibility is where AI can deliver outsized impact.
How has using AI impacted drug development at Formation Bio?
We think about AI in three tiers of impact. Tier 1 is automation—using AI to streamline repetitive, time-consuming tasks like medical writing for protocols, clinical study reports (CSRs), and other documentation. This tier is already creating real efficiency gains across teams. Tier 2 is intelligence augmentation—where AI helps aggregate and synthesize data in ways that make it easier for humans to analyze, interpret, and act on. We're seeing strong results here, especially in early research, trial design, and patient insights. Tier 3 is decision support—AI agents that can make reliable, independent recommendations or decisions in well-defined areas. We're not fully there yet, but we see a clear path and are actively building toward that future.
Right now, most of our work is in Tiers 1 and 2, and it’s already transforming how we operate. But the pace of progress is fast, and I expect material progress in Tier 3 sooner than many might think.
Are there any AI red flags to be aware of?
Yes—AI is a powerful tool, but it's not infallible. One key red flag is when outputs don’t align with your domain knowledge or seem too good to be true. That’s a signal to pause and validate. Always pressure-test AI-generated outputs with subject matter experts and source data when appropriate.
While models have improved and hallucinate less than they used to, they can still produce errors or make assumptions that aren’t grounded in evidence. It’s important not to treat AI as a source of absolute truth, but rather as a partner that needs oversight—especially in high-stakes contexts like drug development.
