Artificial intelligence (AI) will help the drug industry automate aspects of biopharmaceutical process development, but human oversight will always be a necessity, according to new analysis.
A team led by researchers at Imperial College London made the case for the hybrid human-AI approach recently, arguing that while automated “self-driving labs” (SDLs) are an attractive idea, the complex nature of drug production processes makes them impractical.
“The vision of fully autonomous bioprocess development from strain engineering to manufacturing scale remains compelling; yet the current reality points toward a more nuanced trajectory.
“While SDLs have demonstrated remarkable success in chemistry and materials science, bioprocessing faces distinct challenges that, at least initially, point toward an era of hybrid labs,” they write.
“Rather than pursuing full autonomy,” the authors continue, “we foresee that hybrid SDLs, combining AI-driven decision-making with sustained human oversight, represent the most practical near-term trajectory.”
Human oversight
The team found that, although AI systems and lab technologies have advanced to the point where individual processes can be automated, the safety and quality rules that cover drugs and their production will always require human expertise and input.
“The primary bottleneck is no longer hardware maturity: liquid handlers, robotics, and analytical instruments have reached sophisticated levels.
“Instead, we are limited by the orchestration of complex, multiscale workflows under regulatory and safety constraints that demand human oversight,” they write.
The authors predict that, although AI agents and LLM-based tools will play a greater role in process development, engineers will retain the responsibility for things like anomaly detection, planning, and safety-critical decisions.
“This balance is not a compromise but a necessity, reflecting the unique challenges of scaling from high-throughput strain engineering to pilot plants and manufacturing.
“Alongside automation hardware, digital twins, multi-fidelity optimization, and standardized data frameworks provide promising scaffolds to bridge this gap. However, progress will most likely depend less on isolated breakthroughs than on coordinated community action,” they write.
Standardization
Standardization is one example of where “community action” is needed. The development of automated labs will depend on the creation of performance metrics that can be used to monitor and control operations, according to the authors.
“Realizing SDLs for bioprocess engineering will therefore require universal standards for protocols and metadata, benchmarks for scale transfer and sustainable economic models that extend automation.”
They add, “The next frontier is not the replacement of scientists but the careful design of hybrid systems, where human expertise and AI-driven automation complement one another to enable reliable, scalable, and responsible bioprocess innovation.”