An AI Use Case in Pharma Process Optimization
By Katie Anderson, Chief Editor, Pharmaceutical Online
AI is on everyone’s radar in the pharmaceutical industry, but the sheer range of applications can make finding a starting point daunting. This was the case for the Recordati Ireland team, until they saw their yield dip without any real cause. They now had a starting point.
At ISPE’s 2025 Pharma 4.0 Conference, Recordati Ireland’s production coordinator Ross Fitzgerald took attendees through his team’s AI journey, from identifying the problem to using AI data analytics to find a solution.
In 2020, Recordati Ireland saw a decline in yield (-4%) and an increase in yield variability (+/- 7%). Process data was segregated and compliance was never compromised. According to Fitzgerald, “data was rich, but insight was poor.” Like most manufacturing facilities, the company’s data lived in different environments. “It was time to bring our data to life,” noted Fitzgerald.
Bringing Data to Life
And bring the data to life they did by integrating AI into production workflows, supported by IoT and a secure GxP cloud. According to Fitzgerald, they needed to visualize the process. More specifically, they needed to visualize the relationship between process conditions and process outcomes.
Integrating AI into production workflows gave the team real-time visibility of inefficiencies so they could quickly make decisions. The data resulting was contextualized so they could understand the variability.
After three months of data, the team realized that the operators were adjusting the jacket temperatures differently. “This was exactly the insight we needed to identify the golden batch,” noted Fitzgerald.
They studied the data to determine what drying time resulted in the best yield. “We struggled to show the relationship between drying time and yield. Visualizing it gave us the data that helped us make the decision,” explained Fitzgerald.
Improved Yield with No Disruption
After studying the data, the team determined that two hours was the magic drying time to produce the best yield. After applying this, the yield improved 1.5% after just three months. It also reduced cost of goods sold (COGS) by 2%.
Fitzgerald and his team were pleased that they were able to improve their yield by just adjusting their parameters. He noted that no new equipment was needed and no changes were needed for the validated process.
The adjustment to the parameters not only improved yield, but it also strengthened productivity and compliance. On the manufacturing floor, Fitzgerald added that the AI integration also rekindled the focus of the operators. The use of modern tools also improved team motivation. “It accelerated us on our Pharma 4.0 journey,” commented Fitzgerald.
Fitzgerald added that the biggest challenge that remains today was the manual effort the team had to put behind contextualization. “We had to dedicate over 300 hours to manual transcriptions of timestamps of paper batch records into Excel sheets,” he explained. The second challenge, according to Fitzgerald, is that AI is a great tool at finding correlations but should be a starting point. “It isn’t good at chemistry, and it doesn’t’ understand your process,” he continued.
This is the beginning of Recordati Ireland’s journey with AI. “We have some ambitious goals like fully automated data entry via electronic batch records and hopefully proactive process monitoring to prevent and predict deviations in real time. The work done so far is the starting point to step into our next phase of intelligent manufacturing,” explained Fitzgerald.