Experts describe artificial intelligence as one of the most decisive success factors for the media industry, such as in recognition and keywording services for metadata that free up humans to do other things. This talk argues that many products use pre-trained AI models which means the achieved results are tightly correlated to the amount and quality of data the model has been trained on. It presents an alternative approach – 'AI customization' – aiming to provide tools that can be used immediately and without prior technical knowledge. The goal is to let users build their own, individualized AI models and to provide several options for manual training. These customized models could then provide more detailed and higher quality keywording and thus an improved searchability of the media data and allow a wider range of use cases to be covered.