AI is unlikely to take off among CDMOs at the same scale as biopharma companies, insiders say

With AI drumming up interest in drug discovery and R&D, Endpoints News asked insiders how CDMOs are thinking about using AI tools in manufacturing. Their response? AI uptake is stymied by tough regulatory checks, difficulties in gathering data and a high cost burden.

A spokesperson for the FDA’s CDER told Endpoints in an email it has seen limited uses of AI in biopharma manufacturing based on regulatory submissions it has received. 

However, there are still pockets in manufacturing processes that would benefit from AI, such as facility monitoring and speeding up tech transfers, the insiders said.

Justin Butler

Producing pharmaceuticals is a highly regulated process due to the need for consistent drug batches. “Once that manufacturing system is locked, it’s pretty much locked in, and you’re going to want deterministic systems to do what you want them to do and not give leeway to AI,” Eclipse Ventures’ partner Justin Butler said.

He added that a so-called deterministic system will always give the same output, while AI has more variability, especially generative AI. Eclipse has invested in manufacturers such as Cellares and Nucleus RadioPharma.

There are also data considerations. “The biggest challenge for the industry is to integrate new and existing data together in one place,” Lonza’s head of R&D biologics Atul Mohindra said, adding that CDMOs are still typically paper-based, so there is no need to digitize data.

The return on investment is also challenging to demonstrate, dissuading CDMOs from investing heavily into AI, Mohindra said. Employees will need to be trained to be able to interpret data and draw conclusions based on AI output, which can also take time, he added.

AI benefits

Atul Mohindra

But CDMOs should not dismiss AI altogether, as there is a “huge upside” to incorporating AI tools to specific — even if limited — parts of manufacturing, Butler said. “The biggest advantage of using AI is that it allows us to make much better decisions on the back of intelligence,” Mohindra added.

Larger CDMOs might find it easier to use AI because they have more facilities and larger datasets, Mohindra said. But newer, smaller CDMOs will have the advantage of being able to use AI from the outset, he added.

Mohindra said Lonza is using AI to monitor facilities, referred to as “predictive maintenance,” to avoid site shutdowns by monitoring changes and keeping facilities up to regulatory standards. The CDMO also uses AI to speed up the process of tech transfers, he added.

There is an increasing trend of biopharma clients using AI for drug discovery, and protein and gene assembly, Mohindra said. And so manufacturers that offer R&D services will find it beneficial to use AI too, Butler added.

AI can also help improve the reliability of supply chains by predicting future demand for drugs, the FDA spokesperson said.