Role of AI in Pharma Manufacturing: Early Warning Systems, Incremental Automation, and the Path from Paper to Digital
A 30-year pharma veteran discusses why FDA observations keep repeating, why APRs are a ritual rather than an improvement tool, and how companies should actually approach the transition to AI.
Vivek Gera Host
Co-founder · Leucine
Dr. Rajiv Desai
Former EVP & Global Quality Head · Lupin Limited
About this episode
In the first episode of the Leucine Podcast, Vivek Gera sits down with Dr. Rajiv Desai, Former EVP & Global Quality Head at Lupin Limited, with over three decades of experience in quality and manufacturing. The conversation covers why the same FDA observations recur across companies and inspections — because corrective actions are local, not holistic — and why annual product reviews have become a compliance ritual rather than a process improvement engine. Dr. Rajiv Desai argues that the path to AI in pharma is not a leap to full digitisation but a deliberate journey from less paper to paperless, with early warning systems and real-time data interpretation as the most actionable first steps.
Topics
Key takeaways
- FDA 483 observations repeat because companies take localized corrective actions — fixing only the cited area without applying the lesson holistically across all related processes and sites
- Annual product reviews are prepared for regulators, not for continuous process improvement — the data exists but the insight loop back into manufacturing is almost never closed
- The path to AI is not paperless on day one — it starts with less paper, incremental automation of transcription-prone activities, and early warning systems before attempting full digitisation
- Young pharma professionals are more receptive to technology than senior leadership; the real opportunity is pairing their conviction with tools that do not threaten the regulatory filing
- Fear of regulatory non-acceptance is the primary mental block for AI adoption, despite FDA having already approved PAT applications — the door is open, but companies do not test it
- Can we get something called an early warning system — an out-of-trend signal during the process, so we correct before failure rather than investigate after it?
Powered by
LeucineOS
See how an AI-native quality platform unifies QMS, compliance, and manufacturing operations.
Explore LeucineOSSee how Leucine's AI-native platform is transforming pharmaceutical quality and manufacturing operations.
