 We're surrounded by the Internet of Things. Thousands of sensors are laying data from cars, factories and stores. These sensors and the data they produce ensure safe driving, food safety and efficient production. However, there is much more we can do with the large amounts of information that is generated. At IBM Research, we're developing AI technologies to connect and understand data in ways never seen before. We combine machine learning with knowledge graph reasoning to enhance the data with layers of semantic abstraction. This gives us new natural user interfaces that allow us to gain new insights from the data. A lot of the data that's coming from us is IoT based and that's Internet of Things based, so that's machine to machine data. When you get a lot of data thrown out, it ends up being a very difficult task of trying to extract real insight from that data set. The only way we're going to solve this problem is to teach the machines to make sense of the data so we can ask it questions and expect reasonable answers. We developed a knowledge graph for IoT. The system knows all about IoT and understands the meaning of different types of data. It works in a similar way to the way our brains think. Imagine a temperature sensor in a building. An office worker tells the system, hey, it's too hot in here. It analyzes the speech's text and extracts the concepts hot in here. It understands that hot is a concept for heat. It quickly accesses the knowledge graph to find all assets in the area. It then uses AI analytics and sensor data to determine if the temperature is actually too high. It realizes the office area is located close to a window. It checks the illumination sensor, automatic blinds and weather to see if there is high sunlight causing excessive temperature and finds nothing unusual. It tries a different branch in its knowledge graph. It finds an AC unit and uses AI diagnostics to find the heating valve switched on. So we can actually identify the problem and then commission work automatically to the relevant people to go fix that problem. It is a scalable solution that enables IoT to learn behavior to understand operation and to self-diagnose problems while making human machine interaction more natural and intuitive. It's a general purpose artificially intelligent multi-dimensional knowledge graph for IoT. It combines reasoning with machine learning to analyze and understand very large data sets in real time. This enables new insight from complex systems at scale.