“I call it a music concierge, like your music best friend,” said O’Driscoll at VentureBeat’s MobileBeat conference in San Francisco, “an ever-present music concierge in some sort of group environment that could be a group chat in Slack or it could be living in the background in the home as part of a voice in a speaker.”

“When the mood strikes you to find some new music, he’s there or she’s there as an interface layer that can search and return music for you,” he said.

O’Driscoll stopped short of saying that Spotify is developing its own smart music concierge, but in describing a couple of discovery tools he’d like to see, O’Driscoll shared the challenges the company faces with regard to natural language processing — the sort of technology that Spotify may need in order to build a truly autonomous voice or text-based music concierge.

According to O’Driscoll, NLP “is a massive investment to make and I don’t think we are where we need to be with regards to making these compelling experiences for users right now,” O’Driscoll said.

“Do we heavily invest in this? Or do we invest less in the NLP hoping that somebody else cracks it and we build on their NLP layer? These are all decisions that we are trying to make.”