Artificial intelligence is already transforming biotechnology innovation in agrifood, but it isn’t useful everywhere, experts at the Hello Tomorrow Summit in Amsterdam said last week.
Recent years have proved AI systems’ “capacity to start really unraveling mechanisms in causality in biology in a way that we just haven’t been able to do before,” Pae Wu, general partner at the global venture firm SOSV, said during a panel at the two-day gathering.
But while she and other speakers underscored AI’s growing importance in biotech discovery, they were just as quick to point out where the technology might not work so well in the sector.
Skepticism around personalization
One of those areas is in personalizing biotech food and health solutions.
For much of Silicon Valley, agentic AI is the present and future; let an AI chatbot wander freely across your apps and digital activities, and it can start performing complex tasks tailored to your needs. This type of approach might not map so well onto agrifood, the experts at Hello Tomorrow said.
“I don’t think everything should happen on a personal level,” said Annick Verween, head of the Belgian investment group Biotope Ventures. “What is healthy for me probably is also healthy for you,” she said, and even though “some things will not be,” it usually makes more sense to optimize for commonalities when scaling up for mass production. “I do not think that AI necessarily is a tool that we need here.”
SOSV’s Wu said there are situations where AI “may actually help facilitate scale” and others where it can do the reverse. Venture-backed founders “have to really think about what’s going to happen when you’re at scale” as well as what it takes to get there, Wu said. Frequently that’s substantial marketing spend. (“Getting on those podcasts is not free,” she added.)
This can lead to tough decisions, like balancing the extent of a product’s AI-powered design with the expense of launching it on the market. “Operationally, do those costs line up with the level of personalization that we [investors] do require for these solutions?” she said.
Revving the discovery engine
AI tools have been slashing the costs and timelines of biotech discoveries for the last several years, and leaders in the space said this trend is just getting started.
Nora Khaldi, CEO of the Dublin-based peptide developer Nuritas, said artificial intelligence is helping shift the way scientists tackle complex biological systems. “We think, at max, in maybe three, four dimensions. Biology doesn’t think that way,” she said. “There are many, many more dimensions within it, and AI is going to help actually integrate all that—and see it as a system rather than as one target.”
Khaldi credited her company’s AI platform with identifying a bioactive ingredient in fava beans that she said showed better performance in muscle conditioning than conventional milk-derived proteins. It would have taken human researchers millions of years to pinpoint that ingredient with traditional scientific methods, she said.
Other biotech entrepreneurs at the summit showed off AI promising to speed up biotech innovation. Aquit is using the technology to develop alternatives to animal antibiotics, hoping to address growing drug-resistance problems. CEO Daniela Allerbon said the Chilean startup developed an immunostimulant to provide salmon as a feed additive that protects them against a deadly bacterial infection. Germany-based ScreenSys, which competed with Aquit and five other startups in a pitch session at the event, promoted an AI-guided platform to accelerate crop breeding.
But there were signs that the AI frenzy of recent years may be moderating. Several biotech finalists in the pitch round promoted ways of addressing food system challenges at scale that conspicuously didn’t leverage AI—which for a while has often seemed synonymous with scale. And each of the biotech panelists emphasized that human judgment and process (both regulatory and scientific) would continue to set necessary boundaries on innovation.
“We’re not there yet where AI comes up with something and we immediately just take it to market,” Khaldi said, adding that biotechs can’t deploy scientists merely for quality assurance.
When you’re creating a molecule, efficacy is just one piece of the puzzle, she said: “Is it easily formulated? Can I put it into a pasta, for example, and boil the pasta for 12 minutes?” AI can help account for these application-level criteria early on, Khaldi said, but it can’t rewire the development process altogether.
“There are still a lot of proof points that you have to go through,” she said.
