The idea of having recommendations offered to us based upon the actions of those deemed similar to us is not an uncommon one.  It’s a service that Amazon have offered for an age now, and most retailers offer this kind of service now.

It hasn’t really extended to other areas of online life however.  When we look for restaurants, hotels and the like, research shows that we use reviews extensively, but websites haven’t really tapped into this insight to begin offering recommendations based upon the behaviour of those similar to us.

That may be about to change however, thanks to a new app that was launched at TechCrunch Disrupt recently.  IO is a Russian app that aims to learn our preferences and talks to us in real-time to offer suggestions for restaurants in our locality.

The process begins by the user typing in a question into the app.  The app will then respond (verbally), with a conversation then initiated between user and IO, whereby the user responds as they would to a text conversation.

Based upon this, IO can then recommend restaurants to the user, complete with things like opening hours, menu, location and description.  It also offers users recommendations for specific details about the restaurant based upon previous diners experience.  So you can ask it for instance if there is a particular dish worth trying out, or whether it’s a swell place for a date.

Each time the user is asked to feedback on the relevance of the recommendation, which in turn improves the accuracy of further suggestions.  The aim is that the more the app is used, the more accurate each subsequent suggestion becomes.

The app has initially launched in New York City, but there are plans to rapidly expand across major cities throughout Europe.

There have been attempts to do similar things in recent times, with Scouted aiming to provide similar recommendations, but in a more analogue way.  Rather than using AI to guide their choices, they use an army of ‘scouts’ to highlight places they think are cool, and then they recommend these to you based upon your likeness to each scout.

So a curation based approach rather than an AI based one.  Which do you think will produce the best results?