AI is shifting software from a world of measurement to a world of judgment. For decades, we built systems that optimized for consistency because it could be tested and measured. But AI produces qualitative outputs that cannot be fully asserted, forcing teams to reintroduce human evaluation and rethink testing itself.