Researchers are developing systems that use AI agents to design lab setups in simulations, check procedures before use and, in projects that connect them to lab robots and instruments, direct those machines during real experiments.
Nature reported on November 6 that agents can model a lab workspace and check whether its equipment and layout support the work. Nature also reported that agents can help coordinate lab operations. One example, LabBuilder, turns experiment descriptions into structured protocols and 3D lab layouts designed to meet the model's requirements for space, movement and chemical safety.
Its three components create protocols, arrange equipment and assess the resulting layouts. This gives researchers a way to test a setup in a simulation before building or changing a physical lab. Google Research separately describes VERID as a digital twin, or virtual model, that tracks a workcell's equipment and materials so agents can check commands against the lab's actual condition.
Google says its prototypes have linked agents with lab robots in molecular-design experiments, with results feeding back into the process. Google describes this as research in progress, not as a fully autonomous scientist. Together, the work suggests that laboratories may use software to plan and practice experiments and help carry them out.
Scientists would remain responsible for oversight. The potential benefits are faster setup and fewer avoidable errors. But checks in a simulation alone cannot show that a system will work reliably in every real-world laboratory.