As a growing range of complex therapeutic modalities enables the treatment of previously incurable diseases, quality management systems have had to evolve alongside pharmaceutical manufacturing to support these therapies. 

Quality teams manage increasingly large and complex volumes of data tied to advanced therapies — while also navigating evolving regulatory expectations — and new quality management solutions are emerging to support them in their workflows, including artificial intelligence (AI)-based tools.

Global contract development and manufacturing organization (CDMO) Catalent is among the manufacturers investing in AI for quality management, recently introducing Qai, an enterprise platform designed to support quality system processes across its manufacturing network. 

According to the CDMO, Qai is the first AI enterprise solution launched at Catalent which strengthens “quality management system processes — such as deviations and complaints — by harnessing Catalent’s enterprise data to accelerate analysis, root cause identification and corrective and preventive action development, while improving consistency and speed.”

Managing manufacturing complexity

For Catalent, the platform is expected to play a significant role in supporting quality across its gene therapy manufacturing operations. Because gene therapy manufacturing takes place over a period of weeks or months, there is a lot more process data for quality teams to review throughout production.

“Unlike traditional small molecule products that move through manufacturing steps in a matter of hours, gene therapies move through those steps in weeks,”said Charlie Lickfold, Chief Technology Officer at Catalent. “As you’re going through those steps, there;s a huge amount of data that is created and there’s time for quality to process that data, perform testing, and all this needs to be done and understood.”

For quality teams, that means interconnecting larger sets of data through the process.

“That includes analyzing the historical events that have already happened, the procedures, the specifications,” Lickfold noted. “All that data they have to manage as a part of their job, and they need to do it faster than they did before.”

Catalent’s Qai platform is designed to help quality teams navigate these large data sets, serving as what Lickfold describes as “a highly capable assistant” for the CDMO’s quality professionals. Qai was developed using Microsoft AI technologies powered by Microsoft Azure, including Microsoft Foundry with supporting data and analytics capabilities from Microsoft Fabric.

“We’ve learned that the challenge isn’t about generating the information, it’s really helping the experts that are governing these processes quickly find the right information, connect the dots, and make sure they’re making confident quality decisions,” he said. “This is really where we thought AI had a tremendous opportunity to help these quality professionals synthesize all that information much more efficiently while they preserve the rigor, discipline, and accountability that they have.”

Deviations: A new way to investigate

How investigators respond to deviation-related concerns is becoming more streamlined through the implementation of AI-based quality technologies, which has been a primary area of focus of Qai for Catalent.

According to Lickfold, responding to a manufacturing deviation essentially boils down to identifying the issue, resolving it efficiently, and reducing the likelihood of the unexpected event recurring. AI helps investigators move through those steps more efficiently by surfacing relevant historical information early in the investigation. 

“What our investigators will typically do when this whole process starts is they will review previous deviations; they’re going to look at procedures, standards, manufacturing history, and even prior corrective actions that have happened — and where AI is particularly effective is they can assist those quality professionals in identifying the relevant historical example,” he said. “Surfacing those patterns and organizing information for those investigators to really speed up how they do it, which allows our investigators to spend much less time searching and pulling that information together.”

Lickfold noted how the process of classifying a deviation can be much more streamlined with the help of AI technology — by identifying relevant past examples and presenting information for investigators to consider during classification. 

“They really spend their time looking at that data and determining what they think the final classification should be,” he added.

Regulation and the adoption of AI

While AI is taking on a larger role in quality management, Lickfold said final quality decisions still rely on human judgment. 

“In any area where we have a high degree of regulatory concern, we maintain a human-centered approach as we continue to gain confidence and see the standards around the use of AI in our industry evolve,” he commented.

Lickfold also says regulators are placing increasing scrutiny on how manufacturers validate and govern AI, not just within quality systems but other important infrastructure throughout manufacturing.

“I think first and foremost they’re going to want to see what your approach is to how you plan to leverage AI,” he observed. “What sort of governance framework are you wrapping around it? How are you applying historical practices of testing and verifying things are operating as to your expectations, and how are you applying some of those same practices in methods when utilizing AI?”

According to Lickfold, the CDMO decided to incorporate it first in quality management based on an evaluation framework that looked at workflows that were highly digitized, had an abundance of data to work with, value for the business, and ultimately looked at the feasibility to which it could be implemented into their network.

Ultimately, AI should complement process improvement, Lickfold contends. Catalent’s approach begins with simplifying existing workflows before determining where AI can accelerate them. 

“When we look at areas of opportunity within our operations, the first thing we think of and look for is how we can simplify and improve the process itself first — after that, we then look at where AI can accelerate and improve outcomes,” Lickfold said. 

Catalent developed a formal methodology called transactional process improvement with AI, where the CDMO looks at embedding the technology once the process has been improved. “The goal isn’t to try to just use or have more AI, it’s really just ensuring we have better outcomes,” Lickfold concluded.