Tech Explained: Here’s a simplified explanation of the latest technology update around Tech Explained: Agentic AI Tools Could Transform Drug Development in Simple Termsand what it means for users..
Drug discovery is the process of identifying new candidates for treating diseases. Traditionally, the process of drug discovery and development has been labour-intensive—requiring manual review of disease pathways, identification of drug targets, and a trial-and-error approach to discovering compounds with the desired therapeutic effect.
Following this, researchers must conduct rigorous preclinical work, accounting for accurate experimental design, funding considerations, and ethical concerns. The end goal? A novel compound enters first-in-human trials.
Yet, success observed in preclinical research does not always translate to the clinical world, where up to nine out of ten candidates fail. As a result, investigators must begin again.
Agentic AI has the potential to revolutionize this process. Throughout preclinical work and the clinical trial lifecycle, bespoke AI agents can be implemented not only to support automated tasks but to act autonomously: planning, reflecting, and adapting.
Technology Networks spoke with Dr. Claudio D’Ambrosio, chief revenue officer at ConcertAI, about the role of agentic AI in accelerating drug development. After earning his PhD, he explored how the scientific method could turn evidence into outcomes—”reliably, repeatedly, and fast enough to matter”. He now leads cross-functional teams, bridging science, business, and technology, to translate insights into action.
Of the various technologies you have worked with, what intrigued you about AI, and at what point did you realize its potential impact on clinical trials?
Claudio D’Ambrosio, PhD (CD):
What intrigued me about AI wasn’t the buzzword; it was the combination of scale and adaptability. Clinical trials are full of tasks that are intellectually demanding and operationally repetitive. Historically, those tasks didn’t scale because they depended on scarce expertise and manual effort.
The moment the potential of AI clicked for me was when it stopped being a model that predicts and started becoming a system that can act.
There was a shift toward agentic workflows that can coordinate work across data sources and operational systems. That’s the idea behind Accelerated Clinical Trials, where real-world, public, and third-party data is integrated, followed by the use of agentic AI workflows to automate and inject predictive intelligence across the trial lifecycle.
Can you describe some of the challenges in conducting clinical trials and how agentic AI can be implemented to address them?
A few challenges appear in almost every program I’ve seen. These include recruitment difficulties, site-level friction, operational silos, and reactive monitoring—by the time risks are visible, investigators are already behind.
Agentic AI can help because it’s designed to do more than summarize. It can shorten design and testing cycles, execute workflows, monitor signals, and coordinate responses and tasks across systems. Purpose-built assistants can be created to target the biggest bottlenecks, for example, writing assistant agents that reduce document generation time.
For many reasons, including participant recruitment and study length, clinical trials in cancer research focus on later-stage diagnoses. Do you believe that AI could change this, allowing us to focus more on prevention?
Prevention amounts to multiple cycles of innovation based on where we are today, so it will take time. I think AI can help to shift some effort earlier in the disease progression process, as well as in enriching investigational arms with high responders. This won’t be a single leap, but a series of practical changes that make earlier-intervention studies more feasible.
To be successful, two ingredients matter:
- Finding the right cohorts earlier. Real-world, multi-modal datasets can help to identify patients before they become late-stage cases.
- Making long-horizon studies operationally realistic. Agentic AI, coupled with innovations such as liquid biopsies, will reduce issues with patient enrolment, site activation frictions, and monitoring burden so that sponsors can afford to run more ambitious programs and move earlier in the disease continuum.
How could agentic AI specifically contribute to “level three” reasoning and decision-making in clinical trial design, and do you foresee any limitations?
Today, clinical trials are inefficient despite expert humans who can reason at every step. When I think about “level three” reasoning in clinical trial design, I’m thinking about a system that can take an objective and then actively work through the trade-offs to recommend a design that you can defend.
Level three AI reasoning
This describes an advanced stage of AI in which models go beyond basic pattern recognition or text generation, demonstrating the ability to reason, plan, and solve complex, multi-step problems.
Agentic AI is a better fit for that kind of work because it can coordinate and simulate the permutations of multiple steps rather than stopping at a single answer. The result is that a design workflow becomes more like an experiment: you can compare candidate designs, model inclusion and exclusion impacts, and simulate enrolment timelines before the first site ever opens.
The limitations are important to consider. First, reasoning quality is bounded by data quality and alignment; if inputs are incomplete or inconsistently mapped, outputs will reflect this. Integrating and aligning inputs first should reduce that friction, but it doesn’t eliminate the need for data governance and review. Second, regulated work demands traceability—why a recommendation was made and what evidence supported it. Finally, there’s an operational limit: a recommendation that can’t be executed inside real clinical systems and workflows is still just a slide. That’s why the “agentic” part must include workflow execution.
Finally, how do you predict that agentic AI may change the clinical trial landscape in the coming years?
I expect agentic AI to change trials in a very practical way. Across the trial lifecycle, it will reduce the time between noticing a problem and doing something about it.
In the near term, the biggest shift will be upstream. Trial design will become more efficient because teams can validate feasibility earlier and reduce the expensive cycle of amendments. Over time, this will flow downstream into operations. AI-assisted workflows will reduce manual burden, and monitoring will become less retrospective. Instead of waiting for weekly reports, investigators can continuously track performance signals, identify risks, and receive recommended remediation actions.
The net effect is that clinical development starts to look less like a sequence of handoffs and more like an integrated operating system—one where the decisions are evidence-backed, and the busy work is handled by automation so people can focus on judgment.