Manufacturing Intelligence Episode 1 - Part 2: Automating Deviations with AI

According to Rick Jaenisch, senior director of Education, Outreach, and Digital Experience at OpenBiopharma, every large life sciences company is currently adopting either proprietary AI tools or enterprise versions of models like ChatGPT, Gemini, and Claude. He notes that Claude has been particularly popular because it features a model specialized for life sciences that has been utilized by the FDA.

Jaenisch identifies the most significant use case for generative AI in the industry as the management of deviations, specifically by turning traditional paperwork into AI-fillable forms. He explains to Christopher Cole, associate editorial director at PharmTech, that this application can save a writer anywhere from one to four hours per report, which is especially valuable since manual typing is often slowed down by multitasking and distractions.

In more complex investigations, Jaenisch describes how AI assists teams in gathering disparate data and summarizing video transcripts to support root cause analysis. However, he warns against the rise of "work slop," which he defines as low-effort AI output that can easily be identified through pattern recognition. He points out that this "slop" creates a calculable economic loss because reviewers must spend more time on the back end fixing errors than the original writer saved. To combat this, Jaenisch argues that companies must implement training that goes beyond basic "read and understand" instructions.

Finally, regarding manufacturing, Jaenisch observes that data silos and regulations like the EU’s Annex 22—which restricts probabilistic models in GxP environments—remain significant hurdles. While AI is already established in early drug discovery, Jaenisch is looking toward a future in which "AI-native" pharma companies fully integrate these technologies into the production lifecycle.