Building recommender systems that work

Recommender systems are widely used in e-commerce platforms, where users regularly receive unsuitable tips, because they are based on the principle ‘suggest more of the same’. In other words: I buy a couch and get recommended another couch the next time I’m on the website.

The clear problem in this case, is that the average consumer typically does not need a second couch once they’ve already purchased one. While humans can understand this next logical step, a machine recommender system cannot process any of this underlying contextual information so it does not really respond well to individual customer wishes.

As a result of these incompetencies, web store assistants, and recommender systems in knowledge-intensive business processes require different methodologies to provide employees and users with truly relevant information, suggestions, and ‘next best actions’ that are actually evaluated as helpful by the user. Using the same example: a stronger recommendation would mean that since I’ve bought a couch, I should next be recommended a coffee table to complete my living room. A recommender system powered by semantic AI has the capability to deliver these happy customer purchasing experiences. 

In this talk, we will address possible use cases for recommender systems built on semantic AI, discuss scenarios from different industries (retail, manufacturing, software industry and pharma) and look at how recommender systems serve as building blocks to the broader concept of ‘Enterprise 360’.