The Opportunity for AI in Science Isn’t Where You Think
It’s finally time for bio tech
When people think about AI in science, they picture robo-scientists that generate hypotheses, run experiments, and cure cancer. And yes those things will happen eventually — lots of smart people are already working on them.
Basically, people think of building extremely science-y things that don’t exist elsewhere.
The real low hanging fruit is more of an arbitrage opportunity. It's taking products that are already working outside of science and mapping them into scientific workflows. It’s about automating the things around the lab that make it slow, expensive, and frustrating.
This is nothing new. When I worked at ResearchGate and Benchling, we often looked to our analogous products outside of science for inspiration — LinkedIn for RG, Notion/other knowledge management for Benchling.
AI simply makes it 10X easier to apply that thinking to the other areas of science that keep it anchored in the past.
Why Science Workflows Are Ripe for Change
The reality of working in science is that most workflows are slow, fragmented, and painfully manual. That’s a big reason why I left neuro labs for tech. The tools that work elsewhere just don't work in science, and the result is products that feel like they were built when the internet was still a novelty.
Take recruiting as an example. Scientific recruiting requires people with highly specialized backgrounds — unique expertise that’s not visible on LinkedIn or in conventional resumes. It’s more akin to finding open-source developers based on their specific code contributions than traditional recruiting. And so, to find relevant scientists you spend hours combing through publication data or you pay ResearchGate to blast ads to relevant scientists.
AI now makes it possible to sift through data like publication history, protocol expertise, or project contributions as a human would to identify the perfect hire, collaborator, or key opinion leader. I know because we’re currently scoping that agent for a potential customer, and it’s just obvious how much of a game changer AI will be.
The fragmented nature of scientific workflows has also made it hard for technologists to step in. Every lab and company operates differently, creating massive scope creep for anyone trying to build scalable tools. Even killer products like Benchling need to become mind-numbingly big and complex to serve a large enough market to be venture scale. As a result, most scientific software has been built by scientists themselves, leading to hyper-specialized tools that lack the polish and scalability of modern tech.
Why Now: AI Changes the Game
AI is lowering the barriers to building great software for bio. Flexible, adaptive systems can handle the complexity of scientific workflows without requiring fully unique solutions for every version of the problem. That means the number of problems in the crosshairs for serious founders is a lot higher.
Looking for opportunities in science is similar to searching in any highly specialized industry right now. Instead of thinking of companies like Harvey as "legal tech," look at them like "automated document processing and creation for legal." A company like Strange Loop then becomes "automated document processing and creation for accounting."
Once you adopt that frame, you won’t be shocked by news like the group partner at YC leaving to raise 30M for Collate — aka "automated document processing and creation for clinical trial compliance."
Here are a few product areas that are working well outside of science but still need killer scientific versions:
Recruiting (see above). This includes hiring but also KOL finding, collaboration, basically anything where the goal is to find specific scientists.
Sales/Business development. There are probably multiple products here given how much of biotech is driven by BD and acquisitions. An example is something like CPG Radar.
Compliance/regulatory. There are already a few cool examples like Collate and Weave but the market and size of problem is just so massive.
Those are just a few I'm personally excited about. There are many others.
It's Finally Time for Bio Tech
Science is overdue for its leap into the modern age. For too long, we’ve accepted outdated tools and manual workflows as the norm, but AI is finally providing the opportunity to change that. The near-term revolution won’t come from replacing scientists with omniscient systems—it will come from empowering scientific organizations with modern, scalable tools that eliminate inefficiencies and let them focus on the work that matters.
Initially, the companies that build those tools will create custom versions for each customer, mapping AI into existing workflows. And the nature of AI means that there will always be some last mile customization (as I've written about previously). But over time, they will evolve into platforms—streamlined systems that transform how science operates on a broader scale.
If you're working at this intersection of AI and bio or curious about the specific agents we've been building at Robo to help biotech companies get to market faster, reach out. We're incredibly excited to help science move forward, and happy to collaborate with others who feel the same.