One of the technologies we’re using at In The Pocket to make digital products smarter is by enabling them to give advice with the help of machine learning.
Are you into scary movies, but your boyfriend isn’t? Maybe you should check your Netflix account and compare it to his - chances are they aren’t exactly the same. Not only will Netflix recommend different shows, it will also change the thumbnails of the series. That’s true: ‘Stranger Things’ might have a scary picture on your account, while your boyfriend will see one which highlights the eighties vibe in the series.
The time when the supermarket or your favorite clothing store sent all its customers the same discount coupons is over. Nowadays it’s all about individualized offerings: shops or suppliers use machine learning to learn more about their customers, so they can send personalized offers.
Platforms such as Netflix and YouTube have so many content that their users would need at least five lives to watch it all. That’s why they use machine learning to learn as much as possible about their customers, so they can recommend the most interesting content for them. Similar techniques are used by webshops such as Amazon, which recommend the products their customers could be interested in based on their characteristics and shopping behavior.
We trained an AI agent to find its way in a 3D model of our office. Learn more about this project by listening to our podcast with our AI Lead Kenny Helsens and AR Lead Kenny Deriemaeker.
Complex planning problems
When we’re talking about planning, we mean determining a sequence of actions that are known to achieve a particular objective when performed. You can think of it as huge sudoku puzzles, but with real business cases. Forecasting airlines future demand patterns, for instance, and optimizing their pricing based on their operational capacity and market conditions.
Thanks to machine learning, it’s possible to optimize your processes without violating some constraint. A couple of years ago, Goldman Sachs nicely illustrated how you should start doing AI. They mapped out a complete business process and identified 127 steps common across every IPO they managed. They then asked themselves: which steps can be fully automated today? They found that machines, instead of humans, can replace half of all decisions.