Poorly aligned data is typically seen as an obstacle to enterprise AI adoption, but is this the wrong way to look at things?