For the last two years, the enterprise AI conversation has largely revolved around experimentation. Could a model answer customer questions? Could it summarize documents? Could it automate workflows?