It’s such an exciting time. We’re — thankfully — living through this transformation, actively witnessing a profound shift in how researchers and professionals in drug development engage with scientific knowledge.

It’s truly remarkable. The most striking impact is the ability to synthesize knowledge and generate hypotheses more efficiently. Thinking of GenAI as an intellectual catalyst, it’s enabling researchers to connect dots across disparate scientific domains in ways that previously would have taken months, if not years.

This era is all about democratizing complex scientific knowledge. Junior researchers now have faster access to synthesized insights, while senior researchers can work with hypotheses that, in the past, would have taken years to develop. It’s an incredibly exciting time.

I love this question because it really dives into how you integrate AI without disrupting existing workflows. There are three key areas: user-centered design, a data-product mindset, and an entrepreneurial or VC approach.

User-centered design is more than just the user interface in a browser or application. It involves deep ethnographic research and a strong understanding of scientific workflows so that you can identify opportunities to remove friction. This is where true user-centered design really comes into play.

Embracing a data-product mindset is critical moving forward. We’ve long talked about data as an asset, but now it’s a necessity. There’s an urgency around treating data as a core component of success, rather than a passive attempt to extract value. AI simply cannot function, scale, or succeed without high-quality data. Thinking about data as a product, one that is reusable and delivers specific value, enables faster, more consistent AI application development and deployment.

I bring up the entrepreneurial VC mindset for two key reasons: Culture and AI delivery cycles. Culturally, we need to embrace a learning-driven environment where failure is not feared but seen as a pathway to improvement. The focus should be on learning, not failure itself. Creating an environment where AI can truly take hold requires this shift in mindset.

However, people are often hesitant. There’s design fixation — many have built successful careers following certain methods, and AI challenges those perspectives. This is where an entrepreneurial, fast-paced, learning-focused approach accelerates adoption.

The second reason is that AI solutions can’t be delivered like traditional technology solutions, such as ERP systems. AI development cycles are much faster. What you build in the next three to six months might need to be retired in 18 months — whether due to market advancements or the ability to integrate newer technologies.

That’s a 64-million-dollar question. Because you’re dealing with teams that are used to very high precision and scientific excellence, we need to think about reproducibility. We can’t compromise on scientific rigor while adopting AI. The two need to coexist.

So, it’s about finding ways to lower the expectations a little. There’s so much hype around AI, its power, and the idea of designing new drugs from the ground up using AI. That’s exciting, and we’ll get there. Some companies are doing cool things, but a lot of what we really need to adopt is workflow-related, helping people see the value.

You have to have measures and be driven by scientific outcomes – whether that’s time given back, new target identification, or just being specific about what you’re trying to accomplish. But it has to be done in a way that complements scientists. It’s not about AI replacing humans; it’s about how AI can increase scientific creativity and unlock scientists’ workflows.

It takes time, but I see a lot of progress.

I’m super excited about test-time adaptation and long-term reasoning. These are going to be game changers, especially in the complex world of drug discovery. There’s an enormous amount of information to synthesize, and the ability to maintain context across disciplines, along with the memory retention that comes with it, will be transformative. I believe we’re going to see a significant impact early this year, if not by year’s end.

Of course, we can’t have this conversation without discussing agentic AI or AI agents. These autonomous helpers will play a crucial role in augmenting our workforce — helping make decisions, guiding outputs — all while keeping humans in the loop. We’re also going to see increasing democratization in this space. One of the most interesting aspects will be how enterprises approach the development of autonomous agents — whether they allow individuals to build their own or implement them as enterprise-wide systems.

Lastly, both a concern and a point of optimism is the growing challenge of energy limitations. As companies race to train larger models, energy consumption is skyrocketing, and we simply can’t build infrastructure fast enough to keep up. I’m hopeful we’ll see advancements that drive energy efficiency, as this is a critical issue we need to address moving forward.

When we think about AI and pharma, especially in the R&D space, it’s more than just automating science. It’s really about amplifying human scientific creativity—unlocking insights and accelerating drug discovery and development to bring much-needed therapies to patients across the globe.