When Google DeepMind released AlphaFold2 in 2020, it solved a 50-year-old problem in biology: predicting protein structures from amino acid sequences alone. The breakthrough predicted structures for over 200 million proteins, a 1,500-fold increase over what decades of laboratory work had achieved. Yet despite this revolutionary capability, structural biologists didn't disappear from the field. Instead, research from Northwestern University's Kellogg School of Management reveals a more nuanced story about how transformative AI actually reshapes specialized fields.

Did AlphaFold Replace Experimental Biologists?

The initial assumption seemed logical: if an AI could predict protein structures as accurately as expensive laboratory experiments, why would scientists continue doing the slow, costly work? But the data tells a different story. The number of journal articles using traditional experimental methods to determine protein structure has not declined since AlphaFold2's release. Structural biologists continue publishing at the same rate, and remarkably, they're still publishing in top-tier journals like Nature and Science, even though their work is now theoretically substitutable with AI.

Ryan Hill, an assistant professor of strategy at Kellogg, and his collaborator Carolyn Stein of UC Berkeley studied this phenomenon by analyzing decades of protein research data. Their findings suggest that rather than replacement, AlphaFold enabled augmentation. "There's often a lot of complementary insight," Hill explained. "The AI is not perfect. There are sometimes variations of the protein or pieces of the structure that are more difficult for the AI tools to predict. The experimental methods can also have quality issues. Combining insights from the experiments and the AI gives us more confidence that we have the correct protein structure in a way that might matter for downstream research".

How Did AlphaFold Change What Scientists Study?

The most significant shift wasn't the elimination of work, but the expansion of it. Hill and Stein discovered what they call a "floodlight effect": AlphaFold2 dramatically broadened the number of proteins that scientists could feasibly study. Before the AI breakthrough, many proteins lacked solved structures not because scientists weren't interested, but because experimental methods were impractical or too resource-intensive. Some proteins simply weren't worth the years of work and six-figure costs required to determine their structure experimentally.

Consider the example of zebrafish reproduction researchers who had identified a key protein but lacked the expertise to determine its structure experimentally. Previously, they would have had to wait and hope another laboratory made a breakthrough they could build upon. After AlphaFold2's release, they obtained an AI prediction of the protein's structure and used that information to design further experiments about its function. Their research was eventually published in Cell, one of the leading biology journals. This pattern repeated across the field: within a few years of AlphaFold's release, activity on previously unsolved proteins increased substantially.

Ways AI and Experimental Biology Are Combining Forces

• Validation and Refinement: Researchers use AI predictions as starting points, then conduct targeted experiments to validate, refine, or identify edge cases where the AI model struggles, creating a feedback loop that improves both approaches.
• Enabling New Research Areas: Scientists can now pursue research questions that were previously impractical because they required protein structure information that was too expensive or time-consuming to obtain experimentally.
• Accelerating Downstream Discovery: With structures available quickly through AI, researchers can focus their experimental efforts on understanding protein function and behavior, rather than spending years solving structures.

The broader implication is that when a task becomes dramatically cheaper through automation, it doesn't necessarily eliminate the need for human expertise. Instead, it frees humans to tackle harder problems. "This often happens with technological change," Hill noted. "If a task becomes very cheap through automation, that allows people to do a variety of new tasks they couldn't do before".

What Does This Mean for Drug Discovery?

While AlphaFold has transformed structural biology, its impact on drug development remains more limited. Hill and Stein found no significant acceleration in drug discovery rates following AlphaFold2's introduction. The reason is straightforward: protein structure prediction is just one piece of an enormous puzzle. "Even with a very capable machine-learning tool for those structural steps, it's only one piece of a very large puzzle," Hill explained. "Because there are so many bottlenecks, there is no single task that you could fully automate that would have any meaningful impact on the rate of drug discovery".

However, this limitation also presents opportunity. "The good news is it opens up opportunities, because we can divert our human efforts towards some of those bottlenecks, which will hopefully make us more productive," Hill added. "It will have benefits, but I'm not expecting it to happen overnight".

The AlphaFold story carries particular weight given the timing of its release. It emerged in 2020, two years before ChatGPT sparked widespread anxiety about AI replacing skilled workers. For structural biologists, one of the most specialized professions imaginable, the outcome has been neither replacement nor stagnation, but evolution. The field has absorbed a transformative technology and adapted it to expand human capability rather than eliminate human work.

This pattern may offer a preview of how other specialized fields will respond to AI disruption. Rather than wholesale replacement, the most likely outcome appears to be a reconfiguration of work, where automation handles routine tasks and frees experts to focus on problems that require human judgment, creativity, and domain knowledge. For biologists, that means more time understanding how proteins actually work, and less time waiting for the tools to reveal their structures.