The State of AI in Clinical Research: Lessons from the Industry's Leading Voices
Ram Yalamanchili
Article
8 min
read
Across season 2 of our Breaking Protocol podcast, I interviewed clinical research leaders from pharma, biotech, and CROs to understand the state of AI in clinical research today. They spoke to us in separate conversations, from different vantage points: CMOs, CEOs, a biotech founder, a global consulting lead, a big-pharma data and AI head, a VP of clinical operations. They converged on the same answer to the question: where does drug development actually get stuck today?
Almost none of them said discovery.
The through-line of the season is that AI is accelerating drug discovery faster than clinical development can absorb it. Sponsors already have more promising candidates than they can put into trials, which means the industry's binding constraint has moved from finding molecules to running studies. A trial model built on manual coordination does not scale into that, and no amount of additional headcount closes the gap.
Where the season got more interesting is where our guests disagreed, or admitted they had not yet resolved something. The industry has automated before. Several of these leaders lived through EDC and CTMS and watched the promise of technology harden into a new set of constraints. Their skepticism is not about whether AI works. It is about whether the industry will use it to redesign the work or simply to digitize the same inefficiencies faster. And nobody in the season had a settled answer to the harder question underneath: what happens to the business models, org charts, and careers built around the work that AI Teammates now absorb.
Below are the highlights.
Tom Mather, MD - CMO and Board Member at MCAL Therapeutics
Dr. Tom Mather has practiced medicine for over 4 decades, helping shape clinical strategy, trial design, and regulatory engagement. In addition, he now runs Opthafund, which supports investments of promising medical breakthroughs. One trend that stands out to him is that science is becoming exponentially more complex as medicine and technology advance at a rapid pace. Gene and cell therapies and precision medicine are opening doors that we could barely conceive of in decades past. AI and machine intelligence will radically improve how early-stage drugs are evaluated, manufactured, tested, and de-risked.
As a result, the drug development bottleneck that has traditionally sat on the discovery side is now rapidly shifting to clinical development. This means that the current, labor-intensive method to clinical trials is simply not scalable as more candidate drugs are rapidly introduced. That is where AI Teammates can help reinvent the way studies are executed by taking humans out of the cumbersome parts of the process and letting intelligent automation execute those tasks.
George Magrath, MD - CEO at Opus Genetics
Dr. George Magrath is a surgeon who has also led clinical research programs at biotechs and CROs. Currently the CEO of Opus Genetics, he believes that due to AI, undruggable targets will no longer exist in the near future. Already, pharma is dealing with more promising candidates than it can currently initiate through clinical trials. Like Dr. Mather, he sees the constraint moving from discovery to development.
He believes that companies, in particular smaller biotechs with limited funding, should embrace AI-native clinical operations, which automate much of the task-intensive and low-value work. Study teams can now focus on endpoint selection, protocol design, and patient access instead of site setup and quality checks. These types of improvements will allow biotechs such as Opus to run multiple clinical trials in parallel at far lower cost than traditionally possible. The operational scalability also changes the economics of clinical studies for the industry, and allows more drugs to be introduced to patients.
Alex Mok - Founder of Mantra Bio and executive at Enveda
Alex Mok is a biotech founder and operator who argues that the operational complexity of clinical studies is a major barrier to bringing cures to market, sometimes even surpassing the challenges of scientific discovery. Even at major academic medical centers today, studies are hampered by fragmented, manual, and outdated trial operations. He believes that, going forward, AI-native workflows are necessary to bring human and machine intelligence together inside the same system, rather than relying on the current disconnected processes. This will fundamentally rewrite how clinical trials are conducted, allowing humans to focus on patients and on medical judgment. Agentic workflows will lead to faster study velocity and greater capital efficiency. This translates to faster proof of concept and more shots on goal in the industry's quest to accelerate drug development.
Shobhit Shrotriya - Managing Director, Life Sciences R&D Operations at Accenture and Vice Chair of SCDM
Shobhit Shrotriya is a senior industry veteran who has spent decades leading clinical data operations. He has seen how the promise of technology in the past, such as EDC and CTMS, led to automation of activities rather than workflows. The processes moved from paper to pixels, but the fundamental constraints and inefficiencies did not get resolved. For example, a data entry bot could execute its work in a mere three hours, but still required three days to configure per study. The limitations of technology meant that the industry digitized inefficiencies instead of removing them.
With the arrival of AI Teammates, Shobhit believes technology progression is about to shift from simple automation to intelligent orchestration and agentic transformation. But he is clear that the technology is the easier half of the problem. Humans must also be reskilled to have the judgment and ability to oversee and collaborate effectively with AI Teammates, and that is the half the industry has historically underfunded. Capability arrives on a vendor's release schedule; judgment does not. Organizations that buy the first and skip the second will repeat the pattern he has already watched play out once.
Krishna Cheriath - Head of Clinical Research Data and AI at Thermo Fisher Scientific
Krishna Cheriath, Head of Clinical Research Data and AI at Thermo Fisher Scientific, believes that the siloed automation introduced into clinical research in the past 2 decades has hit its natural endpoint and there are no further innovation gains to be made from these. Instead, the industry now needs an AI-first approach based on intelligent workflow automation. Just like Tom Mather and George Magrath, he believes that while drug discovery is advancing at unprecedented rates, clinical research still lags, creating a drug development bottleneck.
He believes that for an organization to actually achieve AI transformation, the key is people in addition to technology. AI fluency must be built into the entire company, and leadership must be fully bought in. He is also candid that this will not be comfortable. Employees and leaders should expect some volatility, which is a natural part of any such transformation: roles change, familiar work disappears, and teams get reorganized around processes that did not exist a year earlier. Leaders who promise a frictionless transition, in his view, lose credibility precisely when they need it most.
Gaurav Bhatnagar - Chief Growth Officer, Tilda Research
Gaurav Bhatnagar is a clinical research industry veteran with a vision to automate much of trial operations. "The work should run itself," is his mantra as he sees a shift happening from AI as a tool to AI as a teammate. But for most companies, building a scalable production product inside a regulated clinical environment has been too complex to manage. This explains why few true agentic AI systems outside of Tilda Research exist in production today.
Gaurav believes that functional AI Teammates execute the work, in collaboration with humans who provide the oversight and judgment. An AI Teammate should have measurable ROI, quality built in by design with published accuracy benchmarks.
Paula Brown Stafford - CEO of Allucent
Paula Brown Stafford is CEO of Allucent and a long-time industry leader who has seen how clinical research has changed from heavily manual to digitized processes. Technology has led to incremental improvements while also introducing challenges of its own. As a CRO veteran, she is particularly interested in the impact of AI Teammates on study costs, and she raises the most uncomfortable question of the season. Much of the CRO model is priced against time and headcount, so automation not only reduces a CRO's costs, it erodes the basis on which the work is billed. When AI Teammates let high-skilled staff automate costly and time-consuming tasks, what impact does that have on the CRO's revenue?
Rather than seeing it as a threat to the CRO's revenue model, she believes that AI Teammates will create new value, and that both sponsors and CROs can benefit from that value creation. Whether the industry actually shares that value in practice is unsettled. Sponsors will expect efficiency gains to show up in their budgets, and how quickly the commercial models adapt to reflect outcomes rather than hours is a question the industry has not yet answered.
Paulius Ojeras - VP of Clinical Operations at Perceive Biotherapeutics
Paulius Ojeras is an industry visionary operator who has fully embraced the concept of AI-native clinical operations and is using AI Teammates to accelerate TMF operations at his company. Like Gaurav, he believes the work should run itself. His criticism of tech vendors is that most of them are trying to automate small fragments of work rather than redesigning operations around entirely new workflows. It's the same mistake companies made when they simply digitized paper processes. AI Teammates, on the other hand, create the opportunity to completely reinvent existing processes and eliminate traditional drug development bottlenecks. He believes that new AI Teammates will invariably lead to changes in clinical development org structures, with new roles and titles emerging as new processes are formed.
What season 2 adds up to
On the diagnosis, the season is close to unanimous. Discovery is accelerating, clinical development is not, and the industry's constraint has moved downstream. That question is effectively settled.
What is not settled is everything that follows from it, and season 2 left three questions open.
The first is whether the industry redesigns the work or automates around it. Shobhit's three-hour bot that took three days to configure and Paulius's critique of fragment-level automation are the same warning from opposite ends of the market. The technology is not what determines the outcome here; the willingness to rebuild a process rather than accelerate it is.
The second is people. Krishna and Shobhit both land on the same point from different organizations: AI fluency and human judgment are the rate limiter, not model capability, and building them is slower, less visible, and harder to fund than buying software. Both are explicit that the transition will be uncomfortable, and that leaders should say so.
The third is who captures the value. Paula's question about CRO revenue is unresolved by design, because the industry has not resolved it. Paulius expects new roles and titles; Paula expects new value, shared somehow between sponsors and CROs. Neither can yet say what the commercial model looks like on the other side.
That is a more honest picture of the state of AI in clinical research than a consensus would be. The bottleneck has moved, the technology to address it exists in production, and the remaining questions are organizational, commercial, and human. Those are the ones worth arguing about and the ones we will keep putting to our guests.
Register Now
