Najat Khan is a formidable leader in the burgeoning AIDD space. Her career exhibits extraordinary imagination and accomplishment. After a PhD in Chemistry at UPenn, she joined BCG, making Partner in an expeditious fashion. She then joined J&J, where she broke ground in a dual leadership role focused on R&D strategy and portfolio and the integration of data science and digital health across the company’s pharmaceutical pipeline. She’s now with TechBio leader Recursion, where she runs both R&D and Commercial and serves on the Board.
Occam had a wide-ranging and richly textured conversation with Najat on the brave, new world being formed by the epic convergence of medical science and artificial intelligence.
I. The Big Picture – Matters of Structure
Occam: Najat, how do you see organizational structure evolving to maximize the union of these two powerful and discrete scientific disciplines?
Khan: In the early days of AIDD, I think many in biopharma viewed data science and, consequently, data scientists as somewhat of an add-on. As a result, data science teams weren’t always as embedded as they could or should have been in the R&D process. That’s starting to change, fortunately, and leaders are increasingly understanding the importance of truly integrating these emerging capabilities and treating data scientists as equal players with research scientists.
In terms of organizational structure, there are typically two organizational extremes: companies where AI/ML capabilities are decentralized – meaning data scientists sit within research teams, with no centralized data science function – and companies where AI/ML capabilities are more centralized from an organizational design perspective, but with ‘dotted lines’ into chemistry, biology, clinical development teams and so forth. There are pros and cons of both models but, either way, embeddedness and empowerment of ‘bilingual talent’ who speak the language of data science and medical science is key to success.
II. The Rules of the Game – Politesse
Occam: What strategic guidance can you provide to new-age leadership?
Khan: In my mind, there are a few guiding leadership principles that will be very important as we enter this new era of AIDD – across both TechBio ventures and in large BioPharmas committed to integrating AI/ML.
(A) Interoperability – A Necessary Equilibrium
Khan: The respective mindsets of computer science and life sciences must integrate with mutual respect. No matter what your organizational structure is, these two perspectives need to work together seamlessly. The natural human instinct is to silo – and view and structure data science and medical science as distinct disciplines. But if you want to solve the hardest problems in R&D – the ones behind industry’s 90% failure rate – you need true end-to-end integration. Both groups must have an equal voice, and leadership has to actively enforce that. The integration of the WHAT and the HOW is incredibly essential.
(B) The Intelligent Bet - The Allocation of Resources
Khan: AIDD is intellectually complex, and you have to maintain a critical mass of the right people with depth in the talent pool. Too often, I see TechBio companies rich with tech talent, but light on biomedical expertise. In pharma, it’s the reverse – plenty of depth on the science side, but not enough in tech. To be successful in AIDD, you need to strike the right balance.
(C) The Cultural Dimension - Tower of Babel or Lingua Franca?
Khan: Success in AIDD requires what I call “bilingualism”—data scientists and life scientists able to speak each other’s language and respect what the other brings. That doesn’t happen by accident. It takes cultural openness, disciplined learning, and deliberate cross-training so teams can adapt to and maximize new ways of working.
I’ve approached this from two different starting points. At J&J, we came from a science-forward vantage point— data scientists learned the problems medical scientists were solving, while medical scientists learned how data science could accelerate their goals. At Recursion, we’re starting from first principles in technology—re-thinking the entire discovery and development process from the ground up—and integrating deep scientific expertise. The goal in both cases is the same: a fully integrated, mutually fluent team capable of solving problems together in ways neither discipline could alone.
HR can help facilitate integration and culture but, ultimately, it’s on leaders from both domains – whether it’s the CSO, CMO, CTO, or therapeutic area heads – to champion integration and bilingualism within their teams. And when things break down, most often because of territoriality, the fix is always the same: go back to the fundamentals. What is the problem we’re trying to solve, and what’s the best way to solve it together? It’s a team sport.
III. Eye on the Prize - The Omega Point
Occam: What is the goal of AIDD?
Khan: For me, AIDD is about fundamentally reimagining how we discover and develop medicines for patients who need them most. Done well, it can shorten timelines, improve success rates, and open the door to treatments that traditional approaches would never have unlocked. The potential is enormous—but it only matters if it delivers real-world impact.
That impact starts with clarity on the problem you’re solving. In biopharma R&D, the goal is to create differentiated medicines—ones with true clinical and commercial advantage—that can meaningfully improve patient lives.
Where I sometimes see misalignment is in how value creation in AIDD is defined. In pharma, there can be a sense of, “We already know how to make medicines—why use AI?” In more tech-native companies, it can be, “We’ve built an incredible algorithm, so the hardest problem has been solved.” But real value comes from applying new tools and approaches in ways that deliver something meaningfully different—whether that’s a medicine that couldn’t have been made before, or a capability that helps a partner solve a problem faster and better than they could before.
When we lose sight of that end goal, it’s often because of avoidable organizational silos or reluctance to share responsibilities. I’ve seen it in companies large and small. The opportunity in front of us is to break down those barriers, combine the strengths of every discipline, and focus our collective energy on what matters most: delivering innovation that changes lives.
IV. Labor Market Forces - A Fashion Cycle
Occam: Can you comment on the current labor economics of AIDD?
Khan: The most valuable talent in AIDD right now are what I call “bilingual” people—those who can integrate across disciplines, think strategically, and execute well. In this field, that often means people with both deep technical skills and real experience in making drugs or solving important health problems. It’s a rarity.
Organizations need to both attract and retain top bilingual talent and develop the people they already have. That means mentoring scientists to understand how to apply AI effectively, and mentoring data scientists to understand the complexities of biology and medicine. Curiosity is critical here—people who have the activation energy to keep learning are the ones who will grow into these multidimensional roles.
LLMs are shifting the baseline for technical skills, so the differentiator will increasingly be those multidisciplinary thinkers who can also strategize and execute at a high level. And we’ll see another evolution soon—teams will need to work effectively alongside agentic AI systems. That’s going to change organizational capabilities and talent needs in the near future.
In the end, the real challenge isn’t just finding brilliant people—it’s building teams that can connect the dots across science, data, and execution, and stay focused on delivering medicines that truly make a difference for patients.
VI. What is AI? - The Pragmatic Test
Occam: We can look at AI/ML as having an algorithm component and a data component. It seems that the quality of data is becoming increasingly important. Training of the model. What is the best way to get quality data? Is it to create your own or to clean up other databases, which you either buy or are made available to you?
Khan: Both matter, but the key is fit-for-purpose data. In AIDD, more data doesn’t automatically mean better data – what matters is whether that data actually answers the biological problem you’re trying to solve.
Yes, great to have large data sets, but first, need to have good quality
For example, if you’re studying why some patients respond to a cancer therapy and others don’t, you don’t need the largest dataset—you need the right dataset. That might mean focusing on a group of patients whose tumors have been deeply profiled so you can identify the biological patterns linked to response. Depending on the therapy, those patterns could be genetic mutations, changes in gene expression, the presence of certain proteins, or other measurable features that help explain differences in how patients respond.
The goal isn’t just volume – it’s quality and precision. Sometimes that means generating your own proprietary datasets. Other times, it’s about curating and refining existing data to make it relevant and reliable for the problem at hand. In the end, the right data, in the right context, is what drives meaningful insights and ultimately better medicines.
VII. A Consolidated Future – The Proof’s in the Pudding
Occam: How do you imagine that AIDD and TechBio will evolve? Do you think the big companies, whether it's GSK or Merck or J&J, will make their way to success, or will they acquire assets and people from the TechBio space? Does TechBio have an independent future? Do you see Pharma doing what they do with AIDD as they did with R&D in biotechnology companies?
Khan: The trajectory will depend on proof—clear examples where AIDD delivers assets that traditional approaches could not. That’s what will drive confidence, investment, and adoption at scale.
I expect we’ll see a few TechBios evolve into fully integrated pharmaceutical companies—building their own pipelines, bringing products to market, and setting new standards for how medicines are discovered and developed. Those will likely be rare, but they’ll set the benchmark for the field.
For others, the path will be more familiar: licensing, co-development deals, strategic alliances, and in some cases acquisition or “acquihire” to bring in talent and capabilities. Ideally, these partnerships will center on truly novel science—drugging the undruggable, unlocking new modalities, or pursuing targets that were previously out of reach. And when TechBios are acquired, they can continue to reshape how large organizations think, operate, and innovate long after the deal is signed.
Either way, the path has been set—these tools and approaches are already being and will continue to be utilized. The common thread is that success will require reimagining drug discovery from first principles—working on the most important problems, with the right data, compute, and integration of disciplines. Whoever can bring those pieces together effectively will lead.
