AI in Clinical Trials: Solving the 86% Failure Rate
Discover how AI is revolutionizing clinical trials, dramatically reducing the 86% failure rate, and accelerating drug discovery. Learn how AI-powered insights are transforming healthcare
Futurist AJ Bubb, founder of MxP Studio, and host of Facing Disruption, bridges people and AI to accelerate innovation and business growth.
The healthcare landscape is undergoing an intensive transformation, with emerging technologies promising to reshape everything from patient care to drug discovery. But beneath the surface of innovation lies a persistent, costly, and deeply human challenge: the staggering failure rate of clinical trials. Imagine investing billions of dollars and countless hours into research, only for nearly nine out of ten initiatives to fall short. This isn’t just an academic statistic; it represents delayed treatments, squandered resources, and ultimately, patients waiting longer for critical breakthroughs. It impacts daily lives, not just in far-off labs, but in every waiting room, every doctor’s office, and every hope for a healthier future. The current system, despite its advancements, is proving unsustainable, demanding a radical rethink fueled by intelligent intervention.
This pressing issue became a central theme in a recent “Facing Disruption” webcast, where our host, AJ, sat down with Dev Roy, CEO of Roartech and IntraIntel AI. Dev, a technology visionary with a background spanning enterprise architecture, government contracting, and deep AI integration, illuminated how artificial intelligence is uniquely positioned to mend the fractured world of clinical research.
His insights, born from his journey from India to leading a cutting-edge AI firm, offered a compelling preview of how AI isn’t just streamlining processes, but fundamentally changing how we approach data, patient engagement, and strategic decision-making in healthcare. The conversation explored the hidden costs of failure, the urgency of technological adoption, and the surprising role of AI as a catalyst for unity in a historically fragmented sector.
Deconstructing Failure: The High Cost of Disconnected Data
Dev Roy didn’t mince words, highlighting an alarming statistic: 86% of clinical trials fail. If you consider that bringing a new drug to market can cost upwards of $2.6 billion, these failures represent an astronomical waste of capital, time, and human potential. So, why are so many trials derailing? The core problem, as Dev articulated, stems from a deeply fragmented system where data and processes exist in isolated silos.
Think about the journey of a clinical trial. It starts with years of painstaking research, often involving scientists sifting through immense volumes of academic papers, a process Dev noted can consume up to 90% of a researcher’s time. This initial research, while foundational, is often disconnected from the subsequent stages. Then comes the complex protocol design, outlining every detail from patient recruitment to data collection. Many trials falter here, with protocols proving impractical or difficult for patients to adhere to. For example, a trial might require a patient to undergo a specific blood test only when they feel unwell, but also prevent them from eating before the test. If a patient feels ill late in the day, the protocol becomes impossible to follow, leading to missed data points or withdrawal. As Dev explained, “each of those components are siloed and it’s like a separate entity. And there is not much connectivity between each of these components to make the overall trial successful.” This lack of a unified thread leads to myriad issues: patient drop-offs due to complex regimens, researchers struggling to correlate disparate data sets, and ultimately, trials that can’t generate statistically significant or reliable outcomes.
The issue isn’t a lack of intent or dedicated people; it’s a structural flaw in how information is managed and leveraged. In the past, data was organized into rigid databases with rows, columns, and primary keys. While efficient for certain tasks, this structure often stripped away the crucial ‘context’ of the information. As Dev put it, “when you store data in a database... we lose a lot of context.” This loss of context is precisely where traditional data management transformed from an “ocean” of information into a “swamp” - a vast, murky repository where insights are buried. AI, particularly the advancements in large language models and contextual engineering, changes this. It allows for the aggregation of immense, diverse datasets, but critically, it can also infer and retain the relationships and nuances - the ‘context’ - that human researchers or older systems might miss. This ability to “connect the dots” across siloed information is what makes AI a game-changer, transforming fragmented data into a cohesive, intelligent narrative that guides the entire clinical trial process far more effectively.
AI to the Rescue: Connecting the Dots with Contextual Intelligence
The solution to the fragmented clinical trial system, according to Dev Roy, lies in AI’s capacity for “context engineering” - the ability to understand and connect disparate pieces of information in a meaningful way. This is a dramatic shift from traditional database systems where data was often decontextualized. Before, a researcher might have mountains of data on patient demographics, drug interactions, and genetic markers, but without a clear framework to link them, crucial insights remained hidden. Here’s how AI is bringing context back and transforming trials:
Streamlining Research & Protocol Design
The initial research phase of a trial often involves researchers poring over thousands of scientific papers. This manual, time-consuming process is now being revolutionized by AI. Dev notes that AI agents can slash research time by 90% because they can process and synthesize vast datasets, identifying relevant studies, historical trial outcomes, and potential drug interactions much faster than humans. But it’s not just speed; it’s about intelligent guidance. An AI platform, customized to a specific trial’s needs, can then leverage this synthesized research to inform the clinical protocol design. This means creating a trial protocol that is not only scientifically sound but also practical and more likely to succeed. Instead of a 60-100 page report taking months to draft, AI can generate highly informed designs in days, if not hours.
Consider the example of a pharmaceutical company developing a new treatment for a rare autoimmune disease. Traditionally, their research team would spend months manually reviewing journals, patient registries, and previous drug failures. With an AI-powered research assistant, they feed in their initial hypotheses. The AI instantly scans millions of papers, identifies relevant genetic markers, highlights successful and unsuccessful approaches in similar conditions, and even suggests potential patient cohorts based on real-world data. It contextualizes this information, summarizing key findings and flagging potential hurdles, enabling the human researchers to focus on critical analysis and innovation rather than exhaustive data retrieval. This accelerates the formulation of a robust and informed trial protocol, built on the most current and comprehensive body of knowledge.
Digital Biomarkers & Enhanced Patient Adherence
One of the persistent challenges in clinical trials is patient adherence. As AJ shared his personal experience with a Long COVID trial, even highly motivated patients can drop out if the protocol is too restrictive or impractical. AI, combined with digital biomarkers, offers a powerful solution. Dev highlighted discussions with therapeutic companies using subtle sensors - worn like an Apple Watch or even embedded - to constantly capture data on a patient’s physical state. This could include sleep patterns, activity levels, heart rate variability, or even muscle responses during therapy.
This continuous, objective data stream allows AI to monitor patient progress and compliance in real-time. If a patient’s physiological markers suggest they’re not following the protocol or experiencing an adverse event, the AI can trigger an alert, prompting intervention. This is far more effective than relying solely on patient self-reporting, which can be unreliable. “It’s not only just how you feel, but we have a complete control over the actual trial and the patient,” Dev explained. This constant feedback loop means issues can be identified and addressed immediately, rather than weeks or months later. It ensures higher data quality and, crucially, keeps patients engaged and supported throughout the trial.
For instance, imagine a diabetes drug trial where patients are required to log blood glucose levels and take medication at specific times. Using a combination of a smart glucose monitor and an app that tracks medication intake, an AI system analyzes the data in real-time. If a patient misses a dose or their blood sugar spikes consistently, the AI detects the deviation. It can then send a personalized reminder through the app, or even alert a care coordinator to check in with the patient, offering support or clarifying instructions. This proactive engagement, driven by continuous digital biomarker data, significantly improves adherence rates compared to traditional methods that might only detect non-compliance during scheduled, periodic check-ups.
AI as the Clinician’s Companion: Precision Healthcare & SaMD
Beyond trial design, AI is emerging as an indispensable companion for clinicians, addressing the very real constraints of time and cognitive load that often lead to suboptimal patient care. Dev Roy discussed how AI can provide “precision-level response” across various aspects of healthcare, moving beyond simple automation to truly augment human decision-making.
Supporting Clinical Decision-Making
Clinicians today are overwhelmed. They face immense pressure from packed schedules, mountains of patient data, and a constantly evolving body of medical knowledge. It’s simply impossible for any human to be aware of every new treatment, every rare disease manifestation, or every subtle drug interaction, especially across specialties. This is where AI excels, acting as an intelligent assistant that synthesizes information relevant to a specific patient’s profile. As AJ noted, “I don’t want to say miss, because this sounds like it’s a mistake on their behalf. I think we have to acknowledge that healthcare providers are very time constrained, cognitively overloaded.” AI helps bridge this gap.
Dev shared the concept of a “Software as a Medical Device” (SaMD) platform, where a medical product, such as a therapeutic device, comes with embedded AI intelligence accessible via a QR code. A doctor or nurse can scan this code, enter a patient’s identifier, and the AI connects to their Electronic Health Records (EHR). It then provides personalized insights: “This product may not be the best use case or using this person may not be the best idea because of these reasons seven years back she had a bad allergy reaction on this particular medication...” This level of personalized, immediately accessible information helps clinicians make more informed decisions, preventing potential adverse reactions or recommending more effective treatment paths that they might otherwise overlook.
Consider a situation involving a patient with non-small cell lung cancer, as highlighted by AJ. In this rapidly advancing field, new precision interventions emerge frequently. An oncologist, even a highly skilled one, might default to traditional chemotherapy simply because they haven’t had time to absorb the latest research on targeted therapies for specific genetic mutations. An AI companion could review the patient’s genetic profile and instantly flag the most cutting-edge, personalized treatments, complete with supporting evidence and potential side effects, thus preventing suboptimal care. This isn’t about replacing the doctor, but providing them with an expert-level, constantly updated knowledge base at their fingertips.
Democratizing “Dr. House” for Rare Diseases
The promise of AI in democratizing medical expertise is perhaps most striking in the realm of rare diseases. Dev Roy brought up the compelling power of AI to tackle conditions that lack proper definition, formalized solutions, or extensive research. He envisioned a global “rare disease platform” where AI could synchronize data from all over the world, providing potential solutions based on similar cases, even if they are isolated incidents in different countries. As AJ playfully remarked, it’s like “democratizing Dr. House.”
This “Dr. House” effect extends beyond rare diseases to addressing inherent biases in healthcare. Clinicians, like all humans, can fall prey to cognitive biases, such as confirmation bias (”I’ve seen this before, it must be that”) or gender bias (e.g., women often reporting struggles to be taken seriously on certain symptoms by male clinicians). AI, by analyzing objective data and providing evidence-based possibilities, can challenge these biases. It doesn’t have preconceived notions; it simply processes information and presents probabilities and potential links. This objective lens helps clinicians consider edge cases, alternative diagnoses, and treatments they might not initially consider, thereby leading to more equitable and precise care.
For example, a patient presents with a constellation of vague symptoms that don’t fit a common diagnosis. Instead of relying on a human doctor’s memory or typical pattern recognition, an AI system, fed with vast amounts of global medical literature and patient data, could identify a few, extremely rare conditions that collectively account for those symptoms. It might point to a specific genetic mutation or an unusual environmental exposure observed in a handful of cases globally. This ability to “think outside the box” or, more accurately, to “think across millions of data points,” elevates diagnostic capabilities for complex and elusive conditions.
The Urgency of Now: Why Waiting for AI is a Business Blunder
Dev Roy’s message regarding AI adoption was unequivocal: the time to act is now, and the cost of waiting is imminent business irrelevance. The pace of change, particularly with AI, is unprecedented, making a wait-and-see approach a perilous strategy, especially for small to mid-sized enterprises (SMEs).
The Disappearance of the Status Quo
Dev painted a vivid picture of the current technological revolution: “The world is moving at a very fast pace at this point of time. It’s like six months back it was something, and now it’s completely different.” He cautioned against the common SME mindset that AI is “for the big companies” or something to adopt “when the product gets a little bit mature.” This thinking is fundamentally flawed because AI isn’t a static tool; it’s an evolving intelligence. As Dev explained, “AI agent is something that doesn’t come as a adulthood right away. It starts as a kid, then toddler, then kid, then a teenager. Then eventually it becomes a grownup man or woman.” This implies that enterprises need to engage with AI early to “train” their agents, allowing them to mature alongside the business needs. Waiting means starting with a “child” AI when competitors already have a “teenager” or an “adult” one.
The consequence of inaction is stark: “Someone with AI capability will replace you. Whatever work you are doing. If you are not bringing AI into it right now and trying to learn with this process, you will miss out.” This isn’t just about efficiency; it’s about competitive survival. Companies that fail to integrate AI risk becoming the “Blockbuster” of their industry - a cautionary tale of an incumbent that saw disruption coming but failed to adapt. For instance, a medium-sized marketing agency that relies on manual content creation and basic analytics will quickly lose ground to a competitor employing AI to generate personalized campaigns, optimize ad spend, and predict customer behavior at a fraction of the cost and time.
Rethinking Organizational Structure and Talent
The integration of AI isn’t just a technical problem; it’s a cultural, process, and business model transformation. Many organizations, Dev noted, still tend to “just throw bodies” at every functional gap they identify. However, forward-thinking companies are now “very focused not to hire more people in those functions. Rather bringing AI to do that end-to-end offering to enhance the capabilities of those functions.” This reflects a strategic shift from merely filling roles to leveraging technology to build inherent capability within the organization.
This shift also profoundly impacts job seekers. Dev shared a striking anecdote from a recent interview: “Do you know how to build an AI agent who can do your job? Because if you do not, if you cannot build an AI agent who will be doing the data analysis job for me. Then you may be irrelevant in next six months.” This is a stark warning that traditional roles focused on repetitive or process-driven tasks are highly susceptible to automation. The new imperative is to enhance one’s own capabilities using AI, not merely to perform tasks that AI can now do more efficiently. For fresh graduates, this means looking beyond conventional academic curricula and actively seeking internships and hands-on experience in applied AI, focusing on problem-solving with AI rather than just theoretical knowledge.
Consider a large financial services institution. Historically, their compliance department might have hired dozens of analysts to manually review transactions for suspicious activity. Now, instead of hiring more analysts to handle increasing transaction volumes, the institution implements an AI-powered fraud detection system. This system can analyze transactions exponentially faster and more accurately, flagging genuine anomalies for human review. The remaining compliance officers are no longer just reviewers; they become experts in configuring and fine-tuning the AI, investigating complex cases that the AI surfaces, and understanding the regulatory implications of the AI’s output. Their role evolves from manual processing to strategic oversight and advanced problem-solving, underpinned by AI. An individual who can train and manage such an AI system is far more valuable than one who can only perform the old manual checks.
Actionable Recommendations for Navigating the AI Storm
The urgency of AI adoption is clear, but how do leaders and emerging professionals translate this into concrete action? Dev Roy offered clear pathways, emphasizing a proactive, value-driven approach.
For Leaders and Decision-Makers:
Roll up Your Sleeves and Educate Yourself: AI is not solely a “technology problem” for your CTO to solve. As Dev stressed, it’s a “culture” and “process” revolution that will impact every facet of your business. Leaders must invest time in understanding AI’s strategic implications. Attend executive workshops, read authoritative research from institutions like MIT and Harvard Business Review, and engage directly with experts. Do not delegate your understanding of AI’s core capabilities and strategic value. For example, instead of just receiving reports, a CEO might participate in a sprint where their team prototypes an AI solution for a specific customer service bottleneck, gaining firsthand insight into its potential and limitations.
Develop a KPI-Driven AI Strategy: Avoid buying “some bunch of tools” in isolation. Your AI strategy must be deeply integrated with your overall business objectives and tied to measurable Key Performance Indicators (KPIs). What specific business problems are you trying to solve? How will AI directly contribute to revenue growth, cost reduction, market share expansion, or improved customer satisfaction? Start with pilot projects of 6-8 weeks, focusing on tangible value. For instance, rather than experimenting aimlessly, a manufacturing executive targets a 15% reduction in machinery downtime by using AI for predictive maintenance, tracking this against historical data and existing maintenance costs. This clearly demonstrates bottom-line impact.
Embrace a Culture of Continuous Learning and Discomfort: The “future-proof” mindset is one of constant adaptation. Your organization should move beyond the comfort of established ways. Encourage experimentation, even if it means some failures. Dev advised, “It’s okay to be a little uncomfortable because unknowingly you are in an uncomfortable zone.” This involves fostering psychological safety for teams to experiment with AI tools and share lessons learned, rather than punishing unsuccessful attempts. Leaders can promote this by publicly championing small AI pilots, celebrating learning outcomes (even from failures), and allocating dedicated time and resources for employees to upskill in AI literacy. McKinsey research consistently shows that companies with strong learning cultures are more agile and resilient to disruption.
For Emerging Professionals and Job Seekers:
Go Beyond the Conventional Curriculum: Your college degree alone may not be enough. The gap between academic offerings and industry demands is widening. Proactively seek opportunities outside of formal education. Look for internships, join open-source AI projects, or participate in hackathons. Intra Intel AI, for example, offers numerous internships. This hands-on experience provides invaluable “real-world” context that formal education often lacks. A student aspiring to be a data analyst, instead of just completing coursework, could intern at a startup using AI to optimize supply chains, learning practical applications of machine learning in a real business environment.
Master the Art of “AI Agent Building” & Value Creation: Your job is no longer just to “do the thing” but to leverage AI to do it better, faster, or cheaper. Dev’s challenge - ”Do you know how to build an AI agent who can do your job?” - is critical. This means shifting your focus from executing tasks to designing and overseeing systems (human-AI partnerships) that deliver superior outcomes. Your value comes from identifying insights and solving problems, not merely processing data. A junior software engineer, rather than just writing code, might learn to use generative AI tools to accelerate code generation and refactoring, focusing their efforts on architectural design, complex problem-solving, and ensuring code quality. This elevates their role from coder to AI-augmented architect.
Cultivate a Network and an Outcome-Driven Mindset: Don’t wait for opportunities to come to you. Actively connect with senior leaders and experts on platforms like LinkedIn. Focus your outreach on the value you can bring, not just the role you’re seeking. “Focus on value. What value I can bring. People will hire you. People will take you for internship,” Dev asserted. This means articulating how you can use AI to solve specific business problems or enhance efficiency, rather than merely listing technical skills. A recent graduate might reach out to a VP of marketing, proposing how they could use AI tools to analyze social media sentiment with greater depth, offering a pilot project that demonstrates clear value rather than just submitting a generic resume.
The Path Forward: Navigating a Unified, AI-Powered Future
The journey through the disruption of AI in healthcare, particularly in clinical trials, paints a clear picture: the future demands unity, adaptation, and an unwavering focus on human outcomes. We’ve seen how AI can mend broken systems by democratizing data, injecting context, and augmenting critical human capabilities, transforming an 86% trial failure rate into a pathway for accelerated discovery and better patient care. From empowering researchers to streamlining protocol design, from enhancing patient adherence with digital biomarkers to serving as an indispensable companion for time-constrained clinicians, AI is not merely a tool; it’s a foundational shift.
Dev Roy’s vision of a “more united world” powered by AI is not just aspirational; it’s practically achievable when we overcome the silos of data, expertise, and mindset. For leaders, this means shedding complacency, actively engaging in understanding AI’s strategic implications, and meticulously tying AI initiatives to measurable business value. For emerging professionals, it demands a bold embrace of continuous learning, a proactive pursuit of hands-on experience, and a relentless focus on creating tangible value within this evolving ecosystem. The “why now” is urgent because the cost of waiting is not merely lagging, but becoming irrelevant. The “how” involves an uncomfortable but necessary journey of learning, adapting, and building human-AI partnerships that prioritize effectiveness and efficiency.
The path forward is complex, marked by challenges in education, regulation, and organizational inertia. But the opportunity - to revolutionize healthcare, to accelerate life-saving treatments, and to enhance human capabilities across industries - is too profound to ignore. By adopting a pragmatic yet optimistic approach, rooted in clear strategy and an active commitment to continuous transformation, executives and professionals alike can not only navigate this AI storm but also emerge as leaders in shaping a more intelligent, connected, and ultimately, healthier future for all.