AI In Clinical Supply – Hype, Hope, And The Path To Predictive Precision

By Rachel Grabenhofer, Chief Editor, Clinical Supply Leader

AI: The Frenemy. On one hand, it’s a Big Brother operative making dubious, black box-concealed decisions – not to mention a job-replacement threat. On the other, it’s a super-powered data retrieval tool and analyst, augmenting human cognition by offloading cumbersome or routine tasks to make room for emotional contemplation and problem-solving.

Grand View Research projects the market for AI productivity tools will reach US ~$36.38 billion by 2033. To enable this growth, vendors are enhancing the multifaceted capabilities of AI tools to improve accuracy, reduce operational effort across functions, automate complex processes to save time and reduce human errors, and empower decision-making with real-time insights. It’s in this latter capacity of decision-making where some industry experts believe AI – agentic AI, in particular – could circumvent a major clinical supply pain point: forecasting.

Inside Agentic AI

Imagine feeding agentic AI every datapoint available throughout the clinical supply chain – from Internet of Things (IoT) device input and shipment updates, to temperature monitoring and more – to predict supply dynamics, recommend adjustments and autonomously take or propose an action where it’s needed the most. That’s what Gunter Van Hoof, vice president of global clinical trial distribution at Marken, proposed during Clinical Trial Supply (CTS) Europe, held in late February 2026.

To frame this proposal, it helps to consider IBM’s description for agentic AI: an AI system that can accomplish a specific goal with limited supervision. Here, agentic refers to the agency of the model to act independently and purposefully. According to IBM, agentic AI consists of a system of machine learning models assigned to specific sub-tasks that mimic human decision-making. These are coordinated through AI to solve problems in real time.

More specifically, agentic AI collects vast amounts of data through sensors, application programming interfaces (APIs), databases, user interactions, etc. It then analyzes the data using natural language processing, computer vision and other AI capabilities, and makes interpretations based on probabilistic models, utility functions and/or machine learning-based reasoning. Lastly, it chooses the best option based on efficiency, accuracy and predicted outcomes, and executes it either by engaging external systems (APIs, data, robots) or making recommendations to users.

A Vision for Proactive Oversight

As GlobalData reported, Van Hoof believes that while predictive analytics and forecasting currently have a place in clinical supply chain oversight, they aren’t sustainable for the long term. In his view, clinical supply chain oversight is moving from how can we react faster? to how can we be proactive? His answer: Agentic AI.

“This will mean that teams can act faster and act in areas where it is most needed,” he said in the report. “The result will be fewer delays in our industry, with greater predictability, proactive exception management, and real-time tracking, so the expected delivery time will always be known.” He noted that moving away from fragmented systems to a more connected one, where data seamlessly flows together, is a better, more sustainable fit for the future.

Retaining the Human Element

This notion of complete automation sounds like the cure to a major clinical supply pain point, right? Not so fast. As Elisha Lowe, Founder and Principal of IRL Life Sciences Partners, reminds us, in a recent Outsourced Pharma (OP) webinar, we can’t lose sight of the human element. After all, there are real people at the receiving end of clinical supply.

“I’m a strong proponent – when it comes to dealing with human life – of there always being a human in the process,” she explained. “AI tools are great, but at the end of the day, they’re probabilistic text editors.” She added that AI is great at recognizing patterns and could be a very useful solution to help manage complicated workflows. “But I would really love to see a human stay in the process.”

Hesitation and a Call for Clarity

Lisamarie Georgen, founder and principal consultant at GuidedPath Consultants, and another OP webinar panelist, also highlighted that there’s hesitation on the part of sponsors to jump aboard the AI bandwagon.  “From what I’ve seen, there’s a lot of hesitation in … making AI more of an active part in running clinical trials, especially along clinical supplies.”

She emphasized that there is industry buzz around using it for supply forecasting, but acknowledged it’s a murky area requiring clarity. “There’s been talk …of having AI for tracking and forecasting…, but I think there are a lot of ‘unknowns,’ and that’s made some of the sponsors kind of pull back from it.” To Georgen, companies are waiting to see what others discover first to determine how well it works and how it might be implemented.

Generative AI’s Need to Feed

Even so, Steven Jacobs, president of Global BioPharma Solutions and another OP panelist, tipped his hat to the progress that’s been made. “It’s amazing, if we actually take a look. You’ve got generative and agentic AI. All of us are very much used to generative AI, which can do some amazing things – if you feed it. It’s like the plant in Little Shop of Horrors (Audrey) – you’ve just got to keep feeding it so grows big and healthy.”

But therein lies the problem, according to Jacobs. “We don’t have a lot to feed it,” he said. “We don’t share information between companies” – in fact, per Jacobs, we’ve even limited the information we share within our own companies.

What’s more, Jacobs reports that today’s generative AIs are built with retrieval augmented generation (RAG) technology, to prevent hallucinations in their responses. But this technology is limited based on what data it can access.

“Retrieval augmented generation now goes out and checks: Is what I told this person correct?” he explains. But because companies have their own various AIs, these don’t get the chance “to go out and look around.” He continues, “That limits the AI, as well, to some really interesting hallucinations.”

Agentic AI Peaks

So for now, Jacobs sees generative AI as a “cool thing,” but he points to agentic AI for the true potential. “Agentic AI gives us the ability to do low-code and no-code programming to automate a lot of our systems,” he explained, noting this is especially useful for clinical supply professionals who use “monstrous” enterprise resource planning (ERP) systems that were originally intended for commercial planning. “Agentic AI can make that a lot easier – but it still requires … folks that really know … the systems.”

All in all, both Van Hoof and Jacobs look to agentic AI as the future for clinical supply planning – and in my view, it’s not a question of if but when. “I think, when I look at forecasting and planning, linking to an IRT or RTSM system, AI shows tremendous promise for the future,” Jacobs added. “All of these are fantastic opportunities for AI – but right now, agentic AI is at the top of the hype curve – so it’s slowly coming down toward reality, and it’ll be interesting to see what happens.”

As agentic AI moves from hype toward practical application, the industry’s challenge will be to weave its strengths into clinical supply workflows without losing the human judgment that matters most — allowing the technology to serve not as a silver bullet, but as a catalyst for a more connected, intelligent, and proactive future.