Take the stages from the previous chapter and shade each one by how much AI actually contributes today. Discovery and design — the work of choosing a biological target and proposing molecules that might act on it — is shaded heavily. Preclinical testing is shaded moderately: AI helps interpret some of the data, but the experiments themselves are done in cells and animals. The three clinical trial phases and regulatory approval are shaded lightly. Post-approval monitoring is lightly shaded as well, since AI can help scan large safety-report databases, but the reports themselves come from doctors and patients.
The pattern is not random. The heavily shaded stages are the ones where the working material is digital: protein structures, chemical structures, and measured properties that can be stored in a database and processed by a computer. The lightly shaded stages are the ones where the working material is a living system or a human being, and the data only exists after someone runs an experiment or treats a patient.
A useful way to read the diagram is as a gradient rather than a set of categories. AI involvement is highest where data is cheapest to obtain and lowest where data is most expensive, and the shading simply makes that gradient visible.