Is Google Duplex ready for Enterprises?
Even ten years ago, this would be stuff of science fiction, or well executed pranks. A machine calls and navigates non-trivial social interactions, like communicating over a bad line to reserve a table.
Naturally, social media is abuzz with predictions of what it means to the way businesses communicate. Pundits are already talking about Google’s grand entry into the enterprise world. The share price has gone up ~10% within a week of Pichai pulling a Jobs on us.
Wish things were that simple. There are three reasons why it is not ready for enterprise CIOs just yet –
- This is very use case specific, and your use cases are not the use cases Google will start with. As Siri, Alexa and Google Voice have shown us, accuracy really dips beyond proscribed use cases. That kind of inaccuracy may be okay for private users, it is not acceptable to your business.
- You don’t have the data to train something like this for your own use cases. (How much data does Deep Learning need?). They have billions of records, approaching a trillion by some estimates. The ROI for a) generating such data for your own use case, b) annotating that data for deep learning to train; and then c) making something meaningful happen; is surprisingly below zero.
- There are so many alternatives that can make things work for you with disruptive ROIs even today.
This is not to take away from Google’s achievement – its simply fantastic. For interactions that are common with first few Google use cases and are not part of your competitive differentiation, it could be a great option. Plus, Google will keep developing and adding more use cases. You must watch this closely.