Most conversations about AI in life sciences manufacturing focus on what the technology can do. Prof Damien Thompson, Professor of Molecular Modelling at the University of Limerick's Bernal Institute and Director of Rinn Pharma and Biopharma, is interested in a more fundamental question: what is the real ceiling? Ahead of his session at the Life Sciences Industry Conference 2026 on Thursday 22nd October at the Crowne Plaza Hotel, Santry, he shared his thoughts.

For Thompson, the defining constraint in 2026 is not compute power. "We are in an inverted bottleneck where compute power is no longer the rate-limiting factor; the real ceiling is our foundational scientific knowledge of biology, chemistry, pharmacology, and process engineering." Large language models accelerated quickly because they trained on a fully understood corpus. Life sciences does not yet have an equivalent. "Biological systems are non-linear, and pharmacological realities are governed by complex mechanisms science has not yet fully mapped." Where AI delivers is where structure exists: manufacturing. "It is our most structured, sensor-rich environment. AI can immediately unlock value in yield optimisation, predictive maintenance, real-time quality control, and supply chain resilience."

The internal conversations required to scale AI come down to four priorities: digital teams accepting that compute does not override GxP requirements; quality and manufacturing teams demanding explainability and model traceability; domain-specific co-development that solves the non-linear physical realities of cell behaviour and process robustness; and federated data infrastructure that allows models to learn across sites while proprietary data stays secure.

His picture of a truly AI-ready site centres on what he calls the human plus AI co-scientist model. "Instead of passive automation, scientists and operators interact with models that propose, test, and challenge hypotheses within defined biological and physical constraints." Three features make it work: closed-loop systems connecting predictive models with automated experimentation, an autonomous feedback cycle from prediction through to improved prediction, and multi-scale integration linking molecular design to continuous manufacturing. "The truly AI-ready site embodies the Rinn Pharma and Biopharma vision, shifting from reactive troubleshooting to real-time predictive assurance and continuous release."

The question every leader should be asking: "Are we generating the high-quality experimental ground truth needed to train and constrain our models, or are we feeding algorithms noisy, incomplete data?" For those still deciding whether to attend in October, his message is direct. "Advanced smart manufacturing cannot be bought off the shelf or built in isolation; it requires multi-year capability building and national infrastructure. Being in the room ensures that upcoming national research initiatives are shaped around the exact operational, yield, and regulatory hurdles your site is managing today."

On what the community gives back, Thompson is equally clear. "It helps us validate research with industry reality and structure impactful partnerships, fostering the public-private model where national funding supports core research talent and testbeds while industry partners contribute data, validation challenges, and facility access. Collectively, these events champion Smart Factories for Ireland."

Full details and registration are available at pharmaawards.ie/conference, and subscribe to the newsletter to stay informed on speaker announcements and agenda updates.