Nick Armstrong, head of AI digital strategy at CAI, a biopharmaceutical consulting and engineering firm, spoke with BioProcess Insider about the current and future applications of AI and how biotechnology companies can align their expectations and best make use of emerging technologies in development and manufacturing. Armstrong is also a member of the International Society for Pharmaceutical Engineering (ISPE), where he serves as co-chair of the new ISPE Artificial Intelligence Community of Practice.
Armstrong recalled his early career at Genentech, when he was first exposed to the critical role of data, design of experiments, and scale-up studies. “I came in as a process development research associate, trying to scale up processes from two liters to 12,000 liters," he said. He later pivoted into what he called the “dark side of GMP” as a reliability engineer where he extensively worked with data as an end user.
Leveraging that experience, he moved to CAI, where he worked for five years running asset management reliability. “I was able to convince CAI that we need to focus on technology specifically,” he said. “They were gracious enough to let me run a group specifically for digital enablement. My job is to work with emerging technologies like AI, digital twin, and augmented reality to get our folks to deliver their projects better, faster, [and] smarter.”
He said that by adopting such technologies, companies enable their employees to focus on higher value activities that require human ingenuity. In the short term, that means using augmented intelligence (rather than artificial intelligence) to expedite the more mindless day-to-day activities that take a lot of time. “Most of these activities will suck the soul out of you,” he said.
Armstrong detailed tasks that need to be done, such as cataloging equipment by identifying construction materials, tag numbers, and component types before manually entering hundreds of pages of data into a spreadsheet. To improve one of those cataloging tasks, he said “We have an application where we [photograph] the nameplate and [the technology] lifts that information off and populates the database, the spreadsheet, whatever output [we] need.”
He said that companies don’t always think about the time lost when they commit skilled workers to menial tasks that can be performed efficiently by augmented intelligence.
But it is the “AI of the future” that has captured the imagination of people not only within the industry and outside of it, with opinions ranging between extremes. “One extreme is from people who have zero trust in AI,” said Armstrong. Such people think “it's going to take over and ruin the human race. Terminator is going to happen [and] civilization’s over.”
Such people may balk at black box AI simulations because they can’t directly see and understand what is happening inside a given model. But Armstrong compares that to the way that scientists use Chinese hamster ovary (CHO) cells. “We don't monitor every metabolic pathway that occurs inside the cell when we're producing it. We know that if we provide this environment and these inputs, these are the expected outputs that we get.” The same is true of AI, which also relies on process validation. “We have a set of inputs that we can control, and we put these inputs into a model, and we have expected outputs, and they line up. We don't necessarily have to know every little thing that's happening inside.”
In the future, AI technologies will enable pharma companies to run simulations that will streamline development processes while also having massive ramifications in applications such as product validation that normally take months or even years to complete. “If we have a fully accurate physics-based simulated model of the facility, I can now run a million different iterations of it to find the one or the few that we're looking for.”
Simulations can also help ease and boost the success rate of technology transfers, which can be very difficult when moving technologies from one facility into another that is not built for it. Simulations can help predict unforeseen difficulties and unintended consequences that are often experienced during tech transfer. Regarding AI’s potential in saving time and money through complex simulations, Armstrong said, “This is the Shangri La that we hope that AI delivers for us.”
But for such potential to turn into reality, more partners must be brought into the fold, and that means more research and lower costs. Armstrong’s team is working with academic institutions such as North Carolina State University to bring down the costs of establishing virtual models and better understand mechanistic models.
It will take years of industrywide effort before AI meets its full potential, but that doesn’t mean it can’t help people today. “You want to win the hearts and minds of an organization?” Armstrong asked. “Start off with the stuff they hate.”