AI In Clinical Supply: Start With The Problem, Not The Technology
By Laura Hay, senior director of global program management, Trax Group | 2025 Winner, everywoman Customer/Passenger (Leader) Award

There is a lot of excitement around AI and what it can do for supply chains. But I think one of the biggest mistakes businesses can make is getting caught up in the technology itself instead of starting with the problem they are trying to solve.
Clinical supply is no exception.
Consider a clinical trial where patient enrollment is higher than expected at one site but slower than anticipated at another. The original supply plan may no longer reflect what is happening on the ground. One location could be at risk of running short of medication while another has more inventory than it needs. If that information is buried across multiple reports and systems, the supply team may not recognize the imbalance until it becomes urgent.
AI could help identify those changes earlier by bringing together relevant enrollment, inventory, shipment, and demand information. But the value is not in using AI for its own sake. It is in giving the team more time to respond before a potential shortage disrupts the trial.
Where Can AI Make A Difference?
I would start by looking at where people are spending their time.
Are teams spending hours pulling reports together, manually analyzing data, or chasing information across different systems? Is that work taking them away from the decisions and problem-solving that could make a bigger difference to the trial?
These are the kinds of areas where AI may be able to help.
For example, a supply team might be reviewing inventory levels, patient enrollment, upcoming site activations, and shipment timelines to determine whether medication will be available where it is needed. An AI-enabled tool could help bring those data points together and flag a potential mismatch between projected demand and available supply.
That alert would give the team a starting point for investigation. They could review whether a site needs additional inventory, whether a shipment should be adjusted, or whether the forecast needs to be updated based on new enrollment information.
Of course, the quality of that recommendation depends on the information available to the system. If data are incomplete, outdated, or disconnected, the output may not reflect the actual situation.
The goal should be to help teams identify relevant changes sooner and spend less time assembling information. That gives them more opportunity to focus on the decisions that require their experience and judgment.
Faster Does Not Automatically Mean Better
If we invest in AI, we need to be able to demonstrate what is actually improving.
Are decisions being made faster? Are teams identifying risks earlier? Are we reducing unnecessary shipments, avoiding excess inventory, or improving the availability of clinical supplies?
Those are much more useful measures than simply looking at how much work the technology can perform.
For example, a team evaluating an AI tool designed to flag potential supply shortages could measure how much earlier risks are identified compared with its existing process. It could also track how often alerts lead to a meaningful intervention, how many urgent shipments are required, and whether inventory levels remain appropriate across trial sites.
The point is to establish a baseline before implementation and compare results over time. A tool that generates more alerts is not necessarily improving operations. The alerts need to be timely, relevant, and useful enough to support better decisions.
I also do not see AI as replacing the need for experienced people. In fact, experience becomes even more important when teams need to evaluate what the technology is telling them, challenge the results when necessary, and decide what action to take.
Imagine an AI tool recommends moving inventory from one country to another because projected demand appears lower in the first location. Before acting, the supply team would need to consider country-specific requirements, product shelf life, shipping timelines, regulatory constraints, and the timing of patient enrollment. A recommendation that looks reasonable based on inventory numbers alone may not work in practice.
The technology can help surface information, but people still need to understand the clinical supply environment and the consequences of a decision. In a clinical trial, the priority is not simply moving inventory more efficiently. It is making sure the right product is available for the patients who need it.
Make AI Useful For The People Doing The Work
Successful adoption also depends on how teams are supported through change. Introducing a new system and expecting everyone to use it effectively right away is not a strategy.
Before implementing an AI tool, I would look at a few fundamentals.
First, is the data reliable? Teams need to understand where the information comes from, how frequently it is updated, and whether different systems are using consistent definitions.
Second, how will the tool fit into existing workflows? An alert is only useful if someone is responsible for reviewing it and knows what to do next. If the technology creates another dashboard that people have to monitor without clear ownership, it may add work instead of reducing it.
Third, what validation and oversight are required? In clinical supply, teams need to understand how a tool reaches its recommendations, what its limitations are, and what level of review is appropriate before an action is taken. The requirements will depend on the intended use of the technology and the systems involved.
Training matters, too. People need to understand why the technology is being introduced, how it affects their day-to-day responsibilities, and when they should question its output. They also need room to ask questions and learn as the process evolves.
I would rather see a team start with one well-defined problem and learn from that experience than introduce a broad AI solution without a clear understanding of what success looks like.
Start Small, Measure What Matters
For clinical supply leaders evaluating AI, I would focus on three things.
- Start with a specific operational problem. Identify a recurring challenge, such as late visibility into inventory imbalances or time spent manually consolidating supply data. Be clear about what needs to improve before selecting a tool.
- Make sure the foundation is in place. Review data quality, system integration, workflow ownership, validation requirements, and staff training before implementation. These fundamentals can determine whether a tool delivers practical value.
- Measure the outcome, not the technology. Establish a baseline and track whether the solution helps teams identify risks earlier, reduce avoidable expedited shipments, manage inventory more effectively, or make better-informed decisions. Use those results to determine whether the approach should be expanded.
AI has real potential to support clinical supply, but adopting the technology and delivering business value are two different things.
For me, the real measure of AI is not how much work it can do. It is how much more valuable it allows people to become.
In clinical supply, that means using technology to support better decisions, anticipate problems, and give teams more time to focus on what matters: delivering the right product to the right patient at the right time.
About The Author:
Laura Hay is senior director of global program management at Trax Group and the 2025 winner of the everywoman Customer/Passenger (Leader) Award. With extensive experience in clinical supply chain and program management, Hay focuses on helping organizations navigate complex global supply environments, improve operational visibility, and build more resilient approaches to clinical trial supply. Her expertise spans clinical supply planning, forecasting, inventory agility, global logistics, and the evolving challenges facing life sciences organizations.