Aside from the already mentioned usage for computational chemistry, there is also potential for all experimental fields, but there it is crucial that the experimentators
provide their data in a standardized form, otherwise the AI's algorithms learn wrong relationships and draw systematically false conclusions.
link their data with suitable metadata, in the best case via a reasonable database. This can be a simple spreadsheet for sample sets of limited complication, for more sophisticated sets a legitimate SQL database will be more helpful.
While I'm not using AI on my data right now, I am eager to keep my data in an "AI-ready" structure so that when we decide to have a bot crawl through our data at some point, maybe a simpler AI of the year 2024 will already find something useful and we won't have to wait for a "careless operator handling" AI of the year 2030.