Matched molecular pair analysis (MMPA) is a well accepted, transparent SAR analysis method that also enables the generation of new molecules to address a biological or physicochemical goal. This makes it an ideal component of a compound discovery artificial intelligence platform. The methods and challenges of supervised and unsupervised MMPA are addressed with the additional issues of conducting analysis on a large scale and automating processes. Extensions of MMPA to apply to potency data via either matched molecular series or exploiting protein structural data are also reviewed. The opportunities for further developments in the field are discussed.