Centralizing AI Strategy, Ensuring Process Integrity, and Balancing Quick Wins with Long-Term Drug Design in Bio/pharma

In this part 2 of a 3-part interview regarding the presentation “The State of AI in Next-Generation R&D” at CPHI Europe 2025, held Oct 28-30 in Frankfurt, Germany, Eva-Maria Hempe, Head of Healthcare & Life Sciences, NVIDIA, explains that the successful integration of Artificial Intelligence (AI) within the bio/pharmaceutical sector requires addressing organizational, technological, and cultural challenges, according to an analysis of strategic imperatives for AI adoption. She notes that the adoption of AI is posited as a fundamental shift that permeates the entire value chain, encompassing applications across research and development, manufacturing, and commercial activities. Her primary recommendation for industry leaders is to solidify AI as a core central business function rather than treating it as a decentralized or non-essential add-on.

According to Hempe, the current barriers to widespread AI implementation are diverse, including pervasive organizational and data silos, outdated legacy IT systems, and fragmented AI strategies. Additionally, she says organizational compute capacity is a significant bottleneck, often manifested as GPU scarcity, which limits project scalability; this issue is frequently linked to a lack of centralized strategic oversight for infrastructure planning. To sustain AI momentum, Hempe advises organizations to strategically balance highly ambitious, long-term goals—such as shifting drug discovery toward intentional design—with short-term "quick win" projects. While transformative drug discovery requires timelines exceeding ten years, immediate impact can be achieved through applications like automated regulatory filing, clinical writing, and optimized recruitment for clinical trials, she adds.

Overcoming these obstacles mandates a significant focus on change management and culture shift, elements often grossly underestimated, according to Hempe. This involves comprehensively educating both staff and leadership on the capabilities and limitations of AI. Furthermore, process integrity is a critical prerequisite for meaningful digitization, she says, summarizing the necessity of focusing on foundational improvements as "if you digitize or standardize or automate broken processes, you get broken automated processes." Successful AI deployment, therefore, relies on improving standardized ontologies, data interoperability, and ensuring robust compliance frameworks, Hempe concludes.

Check out Part 1 and Part 3 of this interview and access all our CPHI Europe coverage!

Transcript

*Editor’s Note: This transcript is a direct, unedited rendering of the original audio/video content. It may contain errors, informal language, or omissions as spoken in the original recording.

 We already talked about silos, namely data silos, but I think there's also a lot of organizational silos. So, I think the big barriers to adoption are organizational silos, legacy IT systems, fragmented AI strategies, and then also change management is really massively underestimated. It's not just you're doing now AI and everybody's going to jump at it.

You have to find the right quick wins, which are helping people to buy into the opportunity. I also actually see is often GPU scarcity. So there is often just not enough compute for people to run what they would like to run. And that ties back to the centralized or the fragmented AI strategy. If you don't have a centralized AI strategy, you don't really have a good overview of what you need in terms of also infrastructure to run your most important AI projects for the organization. So, the first thing that I hope is going to be addressed, I would recommend people to address, is to really have a centralized AI strategy and have that as a core business function and not just an add-on somewhere.

This is just something that permeates across the entire value chain, and we are seeing use cases from research and development all the way to manufacturing and commercial. So this has to be a core central business function.

The second thing, which I already alluded to, is to really balance quick win projects with more ambitious initiatives. We all hope, and we are all working towards, making drug discovery less a serendipitous discovery and more an intentional design. But that's really hard, and the timelines are just 10 years plus. You still have to run through all the steps of clinical trials. Again, AI can help with clinical trials, but in a way, if you take a drug for six weeks, it's just going to take six weeks. Or if it takes two months or three months if you want to show a long-term survival, it just takes time. So this is the long-term parts of the portfolio, but there are so many also quick wins, which are returning impact immediately. They might not be as sexy as solving drug discovery, but with things like clinical writing, automated filing for regulatory, as well as better recruitment for clinical trials, you can win a lot on a very short time scale, and that's going to allow you to sustain the momentum as well for those more long-term and more ambitious initiatives.

And then I think the last thing, which I also said before, is really change management. So it's educating staff, educating leadership. One of the things about AI is that it's at the same time incredibly powerful and incredibly non-powerful. So you need to understand what it can do to basically also be able to judge what you do with the outputs.

Yeah, really driving that change management process, driving the culture shift, driving process improvements as well. We said before garbage in, garbage out. And another of these sort of sayings is that if you digitize or standardize or automate broken processes, you get broken automated processes. So doing all of those things is a real prerequisite to fully tapping into the power of AI. And then of course some of the more technical challenges like data interoperability, standardized ontologies, and then compliance. But there we are also seeing frameworks emerge and we're working with partners on a lot of these fields.