AI is not a single tool applied at a single point. It appears at several distinct stages, and at each one it plays a different role.
At target selection, AI is used to integrate many kinds of evidence — genetic association data, scientific literature, and measurements of gene activity in diseased tissue — to rank which proteins are most likely to be causally involved in a disease. The role here is prioritization: narrowing a long list of plausible targets to a shorter one worth investigating.
At hit finding, AI is used to predict which compounds in a large library are likely to interact with the chosen target, so that laboratory testing can focus on the most promising candidates. This is called virtual screening, and its role is filtering.
At lead optimization, AI predicts properties such as how quickly the body will break the compound down or whether it is likely to be toxic, allowing chemists to choose which variants to synthesize. The role here is prediction of properties that are expensive to measure directly.
In each case the pattern is the same: AI does not replace the experiment. It decides which experiments are worth running. That distinction — proposal versus proof — will matter throughout this course.