After twelve years at the helm of AI-drug developer Recursion, Chris Gibson will hand the CEO baton to Najat Khan, PhD, the company’s chief R&D officer and chief commercial officer, effective January 1st.
In the 18 months since Khan joined Recursion leadership, the Salt Lake City-based company completed its combination with Exscientia, one of the largest M&A events in the AI drug discovery field, followed by workforce reductions and pipeline pruning in 2025. Gibson, who will transition to board chair, stated that Recursion is seeking a leader with a different skillset as the company moves to later-stage clinical programs.
At the core of Recursion’s technology are large phenomic datasets, spanning transcriptomics, proteomics, metabolomics, and more. These “Maps of Biology” create millions of biological relationships by perturbing human cell lines to instill a disease state and testing compounds to reverse the cells to healthy function. In recent years, the company has grown its end-to-end AI drug discovery platform by integrating clinical data to support patient translation. In October, Recursion unveiled a whole-genome map consisting of 46 million microglial cell images to facilitate the identification of new targets in neurodegenerative diseases.
In this interview with Fay Lin, PhD, Senior Editor, Technology at GEN, Khan unpacks realistic timelines for AI-based drugs to make an impact on the clinic, how to balance risk and speed when investing in therapeutic programs, and the future for Recursion under her leadership as CEO.
GEN: What led to the decision to make this leadership transition and what are your top priorities for 2026 for Recursion?
Najat Khan: This has been a very planned transition. 12 years ago when Chris founded the company, all these applications for AI to improve drug discovery and development was an esoteric idea. Kudos to Chris for the work in the early days, reshaping the sector, building the platform and making some of those early bets. I’m excited to have him be the chair of the board where he will continue to be an advisor to me and the broader team.
The next TechBio inflection point will be around how to go from potential to tangible proof points. Gone are the days where an interesting model is sufficient. We need models that improve the medicines that we’re making.
My first 2026 priority will be to continue to harness our platform to show proof of concept through our partnerships with Roche, Genentech, Sanofi and our internal pipeline.
Second, AI is one of the fastest moving spaces. Every day there’s a new algorithm, foundation model, dataset, or consortium being built. At Recursion, we have a fantastic team called the Frontier Hub which led to the partnership with MIT, NVIDIA and Boltz-2. It’s going to be incredibly important to be at the forefront of these advancements.
Third is team and culture. We need multilingual talent that understands biology, chemistry, AI and business leadership. People often ask: “we have the best data scientists and biologists, but where’s the magic?” The magic comes from having a culture where disciplines are treated as equal and integrated in the drug discovery and development process from start to finish.
My last priority is always going to be shareholder value and meaningful impact for patients. We are going to be incredibly disciplined. Our platform generates a lot of programs and ideas. We’re going to continue to be clear about what could be a transformational medicine versus not and be overall good stewards of capital.
While the “AI boom” has hit in recent years, we have not yet seen these blockbuster AI drug proof points. What are realistic timelines for these technological advances to make a difference in the clinic?
Khan: If your ultimate product is a car, you need a drivable road to navigate. In the TechBio space, that road doesn’t exist. The first generation of TechBio companies have focused on building that data and compute foundation. The large language models (LLMs) that we use today are trained on a corpus of data, but knowledge of biology and chemistry is still limited. Most biologists say we only know 10-15% of biology. Imagine driving a car on a road that is only 10-15% constructed. An end-to-end tech stack from biology to clinical development is going to be critical. Clinical success rate is about 10%. If you double that success to 20%, that still means you’ll have two out of eight programs that work. Most companies have a fraction of those successful programs.
The reality is that it’s going to take time. In any industry, you want to look at the green shoots. Some people call it leading indicators. Discovery timelines are getting shorter. The number of compounds being synthesized is one tenth of traditional approaches because we’re doing more screening in silico.
We’re also seeing algorithms develop. When I was at Johnson & Johnson, my team developed an algorithm that could predict a mutation from a histopathology image. People called it “black magic” and “not even real.” Now we see these algorithms are within NCCN guidelines with Breakthrough Device Designation by the FDA. The world is changing and it’s important to acknowledge these early signs.
When selecting therapeutic programs to pursue, how do you achieve the balance between taking risk, i.e. going after the “undruggable,” and prioritizing speed to get important medicines to patients faster?
Khan: AI approach or not, the North Star is whether this drug will be differentiated by the time it hits the market for patients. Is the target unique? Is it something that’s causal in the patient population? Is it a driver of the disease versus a passenger of the disease?
When we design a molecule, it’s important to triage early to frontload the risk so that we both fail and succeed fast, and don’t burn the dollars that it takes to go into the clinic. 70% of the dollars to make a medicine goes to the clinic, not discovery.
Currently, the discovery funnel is a “V” and we would love to get to a “T.” A “V” means you’re starting from a certain number of substrates and there’s waste as you roll down to the product that goes to the market. A “T” means you have a lot of options to choose from early, but then your failure rate is low. You make smart decisions faster, earlier, and cheaper.
You don’t start with a program for the sake of a platform. You start with a program because it can actually create a medicine that has a differentiated value for patients and shareholders. It starts and ends with that.
How can we further address the translation gap which remains a huge bottleneck to success?
Khan: People think about the translational gap as getting into the clinic, but it’s actually the planning approach before you get into the clinic. Connecting multimodal data, from genetics, transcriptomics, to cell image data to patient data and outcomes is really important. It sounds obvious, but that doesn’t always happen today. The second piece is designing the molecule, which requires pre-clinical in vivo testing, and predictions on dose, absorption, Pharmacokinetics (PK) and pharmacodynamics (PD) in humans. I haven’t seen much innovation in that space yet.
One of the reasons innovation is hard is because failures are not published in papers. At Recursion, we have an automated platform where we’re developing our own datasets in our Salt Lake City labs by profiling molecules and measuring multiple ADMET properties to improve prediction capabilities for how drugs will perform in the clinic. We’re also looking at organoids and other models, as you’ve seen the guidance from the FDA earlier this year in reducing reliance on animal models.
The work around digital twins and simulating patient responses is becoming critical. In early development, we’re exploring different patient populations. We can do more work in silico so that early-stage programs are more focused on validation versus exploration.
The value of AI is going to be the compounding impact across the value chain. If you save three weeks in discovery, but can’t recruit properly in development, you haven’t moved the needle. It all comes back to our core thesis of building an end-to-end AI platform.
Which therapeutic areas where you believe would be the highest opportunity for Recursion to tap into the next three to five years?
Khan: The four main areas that drive a huge amount of disease and unmet need are oncology, neuroscience, immunology and rare diseases. We’ve also internally done work in the cardiovascular and metabolic space. Additionally, the way we do drug discovery is unbiased and not based on one disease area or target. We’re conducting large screens, knocking out every gene in the human genome and starting from a much broader landscape.
Practically, we keep a lens as to where the unmet need is going in the next 5-10 years. We have an upcoming clinical readout for Familial Adenomatous Polyposis (FAP), which is a disease that impacts 50,000 patients across the U.S. and EU, but there’s no approved therapy. The standard of care is surgery to remove polyps in your 20s. By the time patients approach 40, most of their colon is removed. We go after first-in-disease and novel targets where success would have a transformational impact on the treatment and lives of those patients.
As you enter your next chapter as Recursion’s CEO, which leadership traits are you most committed to cultivating and exemplifying?
Khan: First is impact. My mom suffers from a challenging neurological disease where there hasn’t been a new therapeutic in two decades. It’s my duty to use the platform and to make a difference because patients are waiting. If you spend a Friday night at home, with friends or in the emergency room, think about that patient. I think about that more often than I admit publicly. That’s why we have to harness these great models, data, and compute to help patients.
Second is integrity. We have to do things the right way, which means being scientifically and ethically rigorous.
Third is being kind. I’ve been doing this for a long time. It’s hard when people don’t believe in what you do or when your cells don’t grow and you’re staying in the lab at 2:00am. I remember being at Johnson & Johnson working on the COVID vaccine in the middle of the night. Leading with empathy and seeing the good in people is incredibly important.
We’re not just making a slight improvement, we’re aiming to make a significant dent in that 90% failure rate. Genetic medicines have a success rate of 20%. My goal is to go much further with what we’re doing in the TechBio space at Recursion.
