From what I can tell, this seems to reduce the cognitive load of a search session way more than using Google (for certain use cases).

The advantages are: fewer clicks/taps, session state is saved on the page and suggestions reduce typing. They seem like minor things but make a huge difference in practice when you are trying to decide between a few options.

That being said, curious to see if the team can get handle the limitations of machine learning and maintain the quality of search as it scales.

If Google works for you, stick with it! Ozlo is for those questions where using Google would make you do all the work - like finding the best place for a burger in SOMA or automatically remembering restaurants you wanted to try.

You can see some added examples in our blog post: https://medium.com/teamozlo/introducing-ozlo-d5cce73d7ba5