AI Might Save Biotech Jobs

Biotech is still a pretty miserable place to be looking for work. Thousands of people have been laid off over the past few years, companies have disappeared, and experienced scientists are competing for openings that used to be much easier to find.

There are some signs that things are getting less bad. BioSpace says job postings in Q2 were up 15% from a year ago. Biotech R&D postings in June were up 42%, and the number of companies announcing layoffs fell from 64 in Q2 last year to 26 this year.

Still, this is very much an employer's market.

Now, there was another number this week that hints at where the jobs could be.

GenScript reported that its AI-enabled drug discovery business doubled year over year for the third consecutive half-year period.

GenScript is interesting here because it isn't really the company making some other AI model; instead, GenScript helps make them, test them, characterize them, and generate the experimental data needed to find out whether the AI was actually right.

And apparently those AI companies are sending them a lot more work. This is where I think the next hiring boom will be.

The usual promise of AI drug discovery is better prediction, fewer experiments, lower costs, and eventually smaller teams. All of that can be true and we could still end up doing far more experiments.

There is an old economic idea called the Jevons paradox. When something becomes much cheaper or more efficient, people sometimes consume more of it, not less.

Computing is a good example. Computers became millions of times more efficient. We did not respond by using less computing. We put computers into our phones, cars, watches, thermostats, and basically everything else.

AI could do something similar to biology.

A scientist who could previously evaluate 100 protein designs might now generate 100,000. Maybe the model filters out 99.9% of them. Great. You still have 100 proteins worth physically testing.

So AI could reduce the number of experiments needed for each idea while increasing the number of ideas enough that total experimental demand goes up. Then take that same logic one level higher.

AWS made it dramatically cheaper to start a software company because startups no longer needed to buy servers and build their own infrastructure. Each company could operate with less capital and fewer people. But that lower barrier also helped create far more software companies.

Something similar could happen in biotech.

A company that once needed $100 million, its own lab, and 100 employees might eventually start with 20 people, AI models, and a network of CROs and automated labs. That sounds terrible for biotech employment if you only look at the company itself.

But if the lower cost means five times as many companies can exist, that also means many times more available jobs.

Those smaller biotechs still need proteins made, compounds synthesized, assays run, animal studies performed, manufacturing developed, and eventually clinical trials. GenScript is already investing heavily in automation and positioning AI-driven drug discovery as a source of growing demand for its protein business.

So the jobs may not disappear so much as move. Maybe fewer of them sit inside the biotech itself, while more sit inside CROs, automated labs, protein manufacturers, assay developers, and all the other infrastructure needed to turn an AI prediction into something real.

None of this means an AI hiring boom is around the corner. GenScript is one company, biotech hiring is only beginning to show signs of improvement, and automation is simultaneously reducing the amount of labor required to run experiments.

But it creates an interesting possibility. AI might make the average biotech company much smaller. And in doing so, it might make the biotech industry much bigger.

GenBio AI released AIDO Cell, a model designed to simulate how human cells respond to genetic and drug perturbations. The system currently models two commonly used cell lines and combines multiple biological data types, including gene expression and protein-level information. GenBio says the goal is to use it for virtual experiments before moving into the lab. [Link]

Bios Life came out of stealth with $25 million and a partnership with Tempus to build an AI-based cancer surveillance platform. The company plans to combine genomic, clinical and longitudinal patient data to update cancer risk over time, initially focusing on people at high risk and cancer survivors. [Link]

The FDA published a discussion paper on how generative-AI medical devices could be evaluated before and after they reach patients. One proposal is a competency-based framework that tests what a system can reliably do, confirms that performance clinically and continues monitoring it after deployment. The agency is now seeking public feedback. [Link]