Interview with Dave Drodge, Group Lead Digital Transformation, Roche
Did you know AI could shrink the production of clinical trial reports from 12 weeks down to 10 minutes? That’s the kind of productivity gain that Dave Drodge, Group Lead Digital Transformation at Swiss Pharma company Roche, is focusing on.
SciencePOD had the opportunity to speak with David, ahead of the Pharma Meets AI Conference 2025, organised by Ubiq from 7-8th April in Barcelona, Spain, Dave will speak on “AI, You, and the Future of Work,” drawing on his contribution to an upcoming book on the subject and reminding us on the importance of the ‘human in the loop’ to maintain accountability and keep control of where we need to be.
The event will focus on topics relevant to the new AI paradigm in the pharmaceutical industry, including AI-Driven Drug Discovery and Development, Personalised Medicine, Regulatory and Ethical Considerations, and Machine Learning in Clinical Trials.
“We don't want to be just the human in the loop, but we want to be human, so we need to define the spaces where we need to be.”
Please tell us about your current role.
One of the key projects that I am focusing on is how we use AI in our daily work and in a more structured way, and really improve how people can be more efficient and have more time to focus on what’s important strategically and focus on what people, uniquely, can bring to the party.
What are you currently using Large Language Model pilots for?
There are three main areas. One involves daily productivity with off-the-shelf AI. The second involves unbundling processes we’ve used previously, and then re-bundling them in a new way where we take better advantage of AI. There’s a good example in pharma, where the time spent on clinical trial reports was reduced from 12 weeks to just 10 minutes. And the last one involves changing business models, which could be huge, although it isn’t applicable in every industry.
Could you provide concrete examples of success stories you will share at the event?
I think the most important thing is to unbundle the tasks that we do in our individual jobs and then figure out how AI can help us to re-bundle them in a more efficient way. This approach allows us to think about the bigger challenges that we could pursue using this technology, so that we do not just increase our productivity incrementally but fundamentally change what we’re doing. For instance, in writing tasks the emphasis shifts away from brainstorming and drafting text to editing the output of AI. One study showed that the “average time taken decreased by 40% and output quality rose by 18%”.
Can you explain the importance of a human in the loop (HITL) in AI?
There are tasks that could be completely performed by AI, in theory, but people prefer the presence of a person, and this can be a legal requirement in certain contexts. For instance, if you have an AI system that gives a recommendation to a clinician or a lawyer, you want that person to have reviewed the relevant information, and also be accountable for their decisions instead of the system just making the decision. Thus, the human in the loop. It’s hard to hold a computer to account, after all.
Based on your existing experience, what are the key lessons for the future?
I think that we need to always be experimenting, as the technology is moving so quickly. There’s a Harvard Business School study looking at where AI or humans are more efficient and effective. Interestingly, it’s not strictly a curve, it’s a jagged frontier, meaning that there are tasks that AI works better at (sometimes surprising) and then tasks that it’s not so good at. We need to figure out what those boundaries are for ourselves. We also don’t want to be just the human in the loop, but we want to be human, so we need to define the spaces where we need to be. For instance, you read some headlines where a competitor has claimed an amazing step forward with AI, but it just doesn’t work in your company, but a less obvious solution is discovered by your peer who is willing to chip away at the problem.