This week, I had the pleasure of attending the DPHARM conference in Philadelphia, where innovative...
The Rise and Inspiration of Agentic AI in Clinical Development
Listen to the podcast here
Agentic AI in Clinical Trials
5:46
The Rise and Inspiration of Agentic AI in Clinical Development
From MaxisAI Inspire 2025 Panel Discussion
BC Consulting Founder and CEO Bryan Clayton recently participated in a panel discussion alongside Abhishek Gupta of BlueRock Therapeutics and Rakesh Maniar of PKR Health and Life Sciences LLC. The event was hosted by MaxisAI and moderated by Nicole Powell.
Cohost: Okay, let's get into this deep dive.
Cohost: We're looking at the Agents of Change panel from Max's AI Inspire 2025.
Cohost: Really interesting discussion hosted by Max's AI.
Host: Yeah, it featured Abhishek Gupta from Blue Rock Therapeutics, Rakesh Maniar from PKR Health and Life Sciences, and Brian Clayton from BC Consulting.
Cohost: And they really jumped into the thick of it, didn't they?
Cohost: Agentic AI in clinical development, the hurdles, the culture, what's actually working.
Host: They did.
Host: And just quickly for anyone maybe less familiar, agentic AI, we're talking about AI systems that can act autonomously towards a goal.
Cohost: Right.
Cohost: So the big question on the table was how to get these tools from just being like a buzzword executives like to actually being part of the day-to-day grind in drug development, safely, reliably.
Host: Exactly.
Host: And Gupta from Blue Rock really hit on that first challenge.
Cohost: Yeah.
Host: The business case.
Cohost: Yeah.
Cohost: Executives are enthusiastic, apparently, but pinning down the ROI, especially for smaller biopharmas, that's tough.
Host: He split it neatly, though, into quick wins, productivity boosts, like competitive intelligence, automating reports, stuff like that.
Host: And then upend, which is more about
Host: Fundamental financial impact.
Host: Think of fraud prevention, bigger strategic shifts.
Cohost: Makes sense.
Cohost: You have to walk before you can run.
Host: Well, precisely.
Host: And that's where Maynard's point about GXP compliance came in so strongly.
Host: Patient safety, data integrity, efficacy, those are paramount.
Cohost: Absolutely critical in this space.
Host: Yeah.
Host: So any AI agent, before it gets near anything impacting patients or core data, needs really solid governance from the get-go.
Host: Plus, upskilling people and being realistic about budgets, prioritizing hard.
Cohost: That governance piece ties into the cultural side, which honestly I thought was maybe the most revealing part of the chat.
Host: Especially Clayton's analogy.
Cohost: Right, the Maslow's hierarchy thing.
Cohost: Leaders are aiming for the top actualization transformation, while the teams using the tools are stuck down at the bottom, worried about basic safety.
Host: That's the core of adoption, isn't it?
Host: If people are worried, will this AI take my job?
Host: Or what if it messes up and I get blamed?
Host: It just won't embrace it.
Cohost: No way.
Cohost: You need that psychological safety to even experiment.
Cohost: Let it untrust it.
Host: And trust is everything with GXP.
Host: Maniar was very clear.
Host: The human in the loop isn't going anywhere soon.
Host: He called Agentic AI a virtual assistant.
Host: Not really a partner yet.
Cohost: A helpful tool, but one that needs oversight.
Host: Definitely.
Host: Every model involved in GXP work has to be completely ready for audits for inspections.
Host: You need the evidence to back it up.
Cohost: So how do you build that trust, overcome that risk aversion?
Host: Well, Gupta suggested starting small, low-risk tasks like, say, transcribing calls autonomously, using the tech for a risk-appropriate purpose, as he put it.
Cohost: OK, to be chosen.
Host: Exactly.
Host: Help teams get past what Clayton called the cultural storm phase, that bit where there's fear and maybe some internal competition.
Cohost: Before you get to stability and actual performance with the new tools.
Host: Yeah.
Cohost: So moving beyond the theory and the culture, what about actual progress?
Cohost: Where is agentic AI making inroads right now?
Host: We had some good concrete examples, mostly in non-GXP areas so far.
Host: Things like calendar scheduling or helping manage supplies for decentralized trials.
Host: Simple but useful stuff.
Cohost: And the learning there?
Host: That it's not just an IT thing.
Host: You need dedicated business resources involved in training these models.
Host: It takes time and operational know-how.
Cohost: OK.
Cohost: Practical realities.
Cohost: Now, Clayton offered some particularly clear examples of where things are working well, didn't he?
Host: He did.
Host: The one about using agentic AI through chat channels like Teams or Slack to handle common questions from clinical sites really stood out.
Cohost: Oh, yeah.
Cohost: Like, what kind of questions?
Host: Things about randomization, supply logistics.
Host: But apparently, the sites often prefer the instant consistent answers from the AI compared to waiting for a human help desk.
Cohost: That's fascinating.
Cohost: Consistency and speed winning out.
Host: It shows acceptance is possible, even preferable sometimes.
Host: And he also mentioned computer vision and imaging.
Host: That's pretty established now, right?
Host: An AI reader working alongside the human radiologist.
Cohost: That's a second pair of eyes.
Cohost: Builds confidence.
Host: Exactly.
Host: It's a model for how we can integrate AI into more complex decisions.
Cohost: OK, so looking ahead, say, five years out, what was the consensus or was there one?
Host: Huh.
Host: Well, opinions varied a bit.
Host: Maynier was perhaps more cautious, predicting we'd still have humans involved, maybe 60% in the loop.
Cohost: Still significant human oversight.
Host: Yeah.
Host: Gupta hoped we'd reach a plateau in understanding, meaning AI just becomes normal, part of the standard toolkit.
Cohost: Less hype, more routine.
Host: Right.
Host: But Brian Clayton, I thought, offered the most forward-looking perspective.
Host: He argued that, yeah, agentic AI will become established, but the next big wave we need to prepare for is already forming.
Host: Which is?
Host: Quantum AI, leveraging that massive leap in computing power, and federated AI.
Cohost: Federated AI, that's about training models without centralizing sensitive data, right?
Host: Exactly.
Host: Training across different data sets, like different hospitals or research centers, without anyone having to share their raw proprietary patient data.
Host: huge for privacy and collaboration.
Cohost: So the journey is, prove the value, ensure the safety and culture, embed the tech, and then get ready for the next technological leap.
Host: That seems to be the trajectory.
Cohost: It really brings it back to that cultural safety point Clayton made, doesn't it?
Cohost: The provocative thought for you, listening, might be, is your organization building that foundation of trust today?
Cohost: Because it sounds like you'll need it to handle what's coming next.
Host: Well put.
Host: From insights on innovation to the human side of AI adoption, the panel at MaxEI A&I Inspire 2025 reflected both the promise and the responsibility of agentic AI in reshaping clinical development.
