Why AI Alone Won’t Transform Bioprocess Manufacturing

Biopharmaceutical manufacturers are generating more process data than ever, but having more data does not necessarily mean having better insight. As real-time analytics, digital manufacturing, and artificial intelligence become increasingly connected, the quality of the underlying data could determine how much value manufacturers ultimately gain from these technologies. Bryan Hassell, founder and CEO of Nirrin Technologies, discussed the developments he expects to have the greatest impact on bioprocess analytics over the next decade.

What will biopharmaceutical manufacturers need to make predictive analytics a reality?

Artificial intelligence is attracting significant attention across biopharmaceutical manufacturing, but the technology itself may not be the most important piece of the equation. According to Hassell, the value of AI will ultimately depend on something more fundamental, the quality and reliability of the process data it receives.

“I think a lot of the focus today is on algorithms, but the challenge is the reliability and consistency of the type of data that's coming in,” Hassell said.

That challenge is likely to become increasingly important as manufacturers adopt more real-time process analytics. Rather than collecting measurements at discrete points, these technologies can bring analytical measurements closer to the process and generate information continuously.

“The ability to generate this information continuously, it will fundamentally change how everybody operates,” Hassell said.

Continuous measurements could also provide the foundation for tighter integration between analytical technologies and digital manufacturing systems. As process data flows into data historians and other digital infrastructure, manufacturers could increasingly use that information to support technologies such as digital twins.

But generating more data will not necessarily translate into better insights. Measurements need to be reliable, consistent, and fit for purpose before they can provide a meaningful foundation for advanced analytics.

“AI is only as powerful as the data that it's built on,” Hassell said.

If process measurements are noisy or inconsistent, even sophisticated AI systems may struggle to identify meaningful patterns or generate useful predictions. Establishing trustworthy data sources could therefore become an increasingly important priority as manufacturers look to apply AI to their operations.

The combination of high-quality process analytics and advanced data science could eventually change what manufacturers expect from process data. Rather than using analytics primarily to understand how a process operates or determine what happened after an event, manufacturers could use increasingly sophisticated models to anticipate what will happen next.

“As pharma starts to get way more trusted process data, they'll move beyond simply understanding how a process works or what happened towards predicting what will happen next,” Hassell said.

That shift toward predictive manufacturing, Hassell suggested, could represent one of the biggest changes in bioprocess analytics over the next decade. The technology may be increasingly sophisticated, but its effectiveness will still depend on the quality of the information underneath it.

Check out part one of this three-part interview here and part two of this series here.